NYTimes:Blogs: Popular classification:: NYTBlogs3_base
bdanalytics
Date: (Mon) Nov 16, 2015
Introduction:
Data: Source: Training: https://www.kaggle.com/c/15-071x-the-analytics-edge-competition-spring-2015/download/NYTimesBlogTrain.csv
New: https://www.kaggle.com/c/15-071x-the-analytics-edge-competition-spring-2015/download/NYTimesBlogTest.csv
Time period:
Synopsis:
Based on analysis utilizing <> techniques,
Summary of key steps & error improvement stats:
Prediction Accuracy Enhancement Options:
- inspect.data chunk:
- For date variables
- Appropriate factors ?
- Different / More last* features ?
- For date variables
- scrub.data chunk:
- transform.data chunk:
- derive features from multiple features
- manage.missing.data chunk:
- Not fill missing vars
- Fill missing numerics with a different algorithm
- Fill missing chars with data based on clusters
- extract.features chunk:
- Text variables: move to date extraction chunk ???
- Mine acronyms
- Mine places
- Text variables: move to date extraction chunk ???
- Review set_global_options chunk after features are finalized
[](.png)
Potential next steps include:
- Organization:
- Categorize by chunk
- Priority criteria:
- Ease of change
- Impacts report
- Cleans innards
- Bug report
- all chunks:
- at chunk-end rm(!glb_)
- manage.missing.data chunk:
- cleaner way to manage re-splitting of training vs. new entity
- extract.features chunk:
- Add n-grams for glbFeatsText
- “RTextTools”, “tau”, “RWeka”, and “textcat” packages
- Convert user-specified mutate code to config specs
- Add n-grams for glbFeatsText
- fit.models chunk:
- Prediction accuracy scatter graph:
- Add tiles (raw vs. PCA)
- Use shiny for drop-down of “important” features
Use plot.ly for interactive plots ?
- Change .fit suffix of model metrics to .mdl if it’s data independent (e.g. AIC, Adj.R.Squared - is it truly data independent ?, etc.)
- move model_type parameter to myfit_mdl before indep_vars_vctr (keep all model_* together)
- create a custom model for rpart that has minbucket as a tuning parameter
varImp for randomForest crashes in caret version:6.0.41 -> submit bug report
- Probability handling for multinomials vs. desired binomial outcome
- ROCR currently supports only evaluation of binary classification tasks (version 1.0.7)
extensions toward multiclass classification are scheduled for the next release
- Skip trControl.method=“cv” for dummy classifier ?
- Add custom model to caret for a dummy (baseline) classifier (binomial & multinomial) that generates proba/outcomes which mimics the freq distribution of glb_rsp_var values; Right now glb_dmy_glm_mdl always generates most frequent outcome in training data
glm_dmy_mdl should use the same method as glm_sel_mdl until custom dummy classifer is implemented
- fit.all.training chunk:
- myplot_prediction_classification: displays ‘x’ instead of ‘+’ when there are no prediction errors
- Compare glb_sel_mdl vs. glb_fin_mdl:
- varImp
- Prediction differences (shd be minimal ?)
- Move glb_analytics_diag_plots to mydsutils.R: (+) Easier to debug (-) Too many glb vars used
- Add print(ggplot.petrinet(glb_analytics_pn) + coord_flip()) at the end of every major chunk
- Parameterize glb_analytics_pn
- Move glb_impute_missing_data to mydsutils.R: (-) Too many glb vars used; glb_<>_df reassigned
- Replicate myfit_mdl_classification features in myfit_mdl_regression
- Do non-glm methods handle interaction terms ?
- f-score computation for classifiers should be summation across outcomes (not just the desired one ?)
- Add accuracy computation to glb_dmy_mdl in predict.data.new chunk
- Why does splitting fit.data.training.all chunk into separate chunks add an overhead of ~30 secs ? It’s not rbind b/c other chunks have lower elapsed time. Is it the number of plots ?
- Incorporate code chunks in print_sessionInfo
- Test against
- projects in github.com/bdanalytics
- lectures in jhu-datascience track
Analysis:
rm(list = ls())
set.seed(12345)
options(stringsAsFactors = FALSE)
source("~/Dropbox/datascience/R/myscript.R")
source("~/Dropbox/datascience/R/mydsutils.R")## Loading required package: caret
## Loading required package: lattice
## Loading required package: ggplot2source("~/Dropbox/datascience/R/myplot.R")
source("~/Dropbox/datascience/R/mypetrinet.R")
source("~/Dropbox/datascience/R/myplclust.R")
source("~/Dropbox/datascience/R/mytm.R")
# Gather all package requirements here
suppressPackageStartupMessages(require(doMC))
registerDoMC(6) # # of cores on machine - 2
suppressPackageStartupMessages(require(caret))
#source("dbgcaret.R")
#packageVersion("snow")
#require(sos); findFn("cosine", maxPages=2, sortby="MaxScore")
# Analysis control global variables
# Inputs
glb_trnng_url <- "https://www.kaggle.com/c/15-071x-the-analytics-edge-competition-spring-2015/download/NYTimesBlogTrain.csv"
glb_newdt_url <- "https://www.kaggle.com/c/15-071x-the-analytics-edge-competition-spring-2015/download/NYTimesBlogTest.csv"
glbInpMerge <- NULL #: default
# list(fnames = c("<fname1>", "<fname2>")) # files will be concatenated
glb_is_separate_newobs_dataset <- TRUE # or TRUE
glb_split_entity_newobs_datasets <- FALSE # select from c(FALSE, TRUE)
glb_split_newdata_method <- NULL # select from c(NULL, "condition", "sample", "copy")
glb_split_newdata_condition <- NULL # or "is.na(<var>)"; "<var> <condition_operator> <value>"
glb_split_newdata_size_ratio <- 0.3 # > 0 & < 1
glb_split_sample.seed <- 123 # or any integer
glbObsDropCondition <- NULL # : default
# "<condition>" # use | & ; NOT || &&
#parse(text=glbObsDropCondition)
#subset(glbObsAll, .grpid %in% c(31))
glb_obs_repartition_train_condition <- NULL # : default
# "<condition>"
glb_max_fitobs <- NULL # or any integer
glb_is_regression <- FALSE; glb_is_classification <- !glb_is_regression;
glb_is_binomial <- TRUE # or TRUE or FALSE
glb_rsp_var_raw <- "Popular"
# for classification, the response variable has to be a factor
glb_rsp_var <- "Popular.fctr" # glb_rsp_var_raw # or "Popular.fctr"
# if the response factor is based on numbers/logicals e.g (0/1 OR TRUE/FALSE vs. "A"/"B"),
# or contains spaces (e.g. "Not in Labor Force")
# caret predict(..., type="prob") crashes
glb_map_rsp_raw_to_var <- #NULL
function(raw) {
# return(raw ^ 0.5)
# return(log(1 + raw))
# return(log10(raw))
# return(exp(-raw / 2))
ret_vals <- rep_len(NA, length(raw)); ret_vals[!is.na(raw)] <- ifelse(raw[!is.na(raw)] == 1, "Y", "N"); return(relevel(as.factor(ret_vals), ref="N"))
# #as.factor(paste0("B", raw))
# #as.factor(gsub(" ", "\\.", raw))
}
glb_map_rsp_raw_to_var(tst <- c(NA, 0, 1)) ## [1] <NA> N Y
## Levels: N Y# glb_map_rsp_raw_to_var(tst <- c(NA, 0, 2.99, 280.50, 1000.00))
glb_map_rsp_var_to_raw <- #NULL
function(var) {
# return(var ^ 2.0)
# return(exp(var))
# return(10 ^ var)
# return(-log(var) * 2)
as.numeric(var) - 1
# gsub("\\.", " ", levels(var)[as.numeric(var)])
# c("<=50K", " >50K")[as.numeric(var)]
# c(FALSE, TRUE)[as.numeric(var)]
}
glb_map_rsp_var_to_raw(glb_map_rsp_raw_to_var(tst))## [1] NA 0 1if ((glb_rsp_var != glb_rsp_var_raw) && is.null(glb_map_rsp_raw_to_var))
stop("glb_map_rsp_raw_to_var function expected")
glb_rsp_var_out <- paste0(glb_rsp_var, ".predict.") # mdl_id is appended later
# List info gathered for various columns
# <col_name>: <description>; <notes>
# NewsDesk = the New York Times desk that produced the story (Business, Culture, Foreign, etc.)
# SectionName = the section the article appeared in (Opinion, Arts, Technology, etc.)
# SubsectionName = the subsection the article appeared in (Education, Small Business, Room for Debate, etc.)
# Headline = the title of the article
# Snippet = a small portion of the article text
# Abstract = a summary of the blog article, written by the New York Times
# WordCount = the number of words in the article
# PubDate = the publication date, in the format "Year-Month-Day Hour:Minute:Second"
# UniqueID = a unique identifier for each article
# If multiple vars are parts of id, consider concatenating them to create one id var
# If glb_id_var == NULL, ".rownames <- row.names()" is the default
# User-specified exclusions
glbFeatsExclude <- c(NULL
# Feats that shd be excluded due to known causation by prediction variable
# , "<feat1", "<feat2>"
# Feats that are linear combinations (alias in glm)
# Feature-engineering phase -> start by excluding all features except id & category & work each one in
, "NewsDesk", "SectionName", "SubsectionName"
, "WordCount" # Feature Engineering done with prior features
, "Headline", "Snippet", "Abstract", "PubDate"
)
if (glb_rsp_var_raw != glb_rsp_var)
glbFeatsExclude <- union(glbFeatsExclude, glb_rsp_var_raw)
glbFeatsInteractionOnly <- list()
#glbFeatsInteractionOnly[["carrier.fctr"]] <- "cellular.fctr"
# currently does not handle more than 1 column; consider concatenating multiple columns
glb_id_var <- c(NULL
, "UniqueID")
glb_category_var <- "NDSSName.my.fctr" # choose from c(NULL : default, "<category>")
glb_drop_vars <- c(NULL
# , "<feat1>", "<feat2>"
)
glb_map_vars <- NULL # or c("<var1>", "<var2>")
glb_map_urls <- list();
# glb_map_urls[["<var1>"]] <- "<var1.url>"
glb_assign_pairs_lst <- NULL;
# glb_assign_pairs_lst[["<var1>"]] <- list(from=c(NA),
# to=c("NA.my"))
glb_assign_vars <- names(glb_assign_pairs_lst)
# Derived features; Use this mechanism to cleanse data ??? Cons: Data duplication ???
glbFeatsDerive <- list();
# Add logs of numerics that are not distributed normally
# Derive & keep multiple transformations of the same feature, if normality is hard to achieve with just one transformation
# Right skew: logp1; sqrt; ^ 1/3; logp1(logp1); log10; exp(-<feat>/constant)
# glbFeatsDerive[["<feat.my.sfx>"]] <- list(
# character
# mapfn = function(Week) { return(substr(Week, 1, 10)) }
#print(mod_raw <- grep(""", glbObsAll[, txt_var], value = TRUE))
#print(mod_raw <- glbObsAll[c(88,187,280,1040,1098), txt_var])
#print(mod_raw <- glbObsAll[sel_obs(list(descr.my.contains="\\bdoes( +)not\\b")), glbFeatsText])
#print(mod_raw <- glbObsAll[sel_obs(list(descr.my.contains="\\bipad [[:digit:]]\\b")), glbFeatsText][01:10])
#print(mod_raw <- glbObsAll[sel_obs(list(descr.my.contains="pad mini")), glbFeatsText][11:20])
#print(mod_raw <- glbObsAll[sel_obs(list(descr.my.contains="pad mini")), glbFeatsText][21:30])
#print(mod_raw <- glbObsAll[sel_obs(list(descr.my.contains="pad mini")), glbFeatsText][31:40])
#glbObsAll[which(glb_post_stop_words_terms_mtrx_lst[[txt_var]][, subset(glb_post_stop_words_terms_df_lst[[txt_var]], term %in% c("conditionminimal"))$pos] > 0), "description"]
# mapfn = function(description) { mod_raw <- description;
# This is here because it does not work if it's in txt_map_filename
# mod_raw <- gsub(paste0(c("\n", "\211", "\235", "\317", "\333"), collapse = "|"), " ", mod_raw)
# Don't parse for "." because of ".com"; use customized gsub for that text
# mod_raw <- gsub("(\\w)(!|\\*|,|-|/)(\\w)", "\\1\\2 \\3", mod_raw);
# return(mod_raw) }
# numeric
# mapfn = function(Rasmussen) { return(ifelse(sign(Rasmussen) >= 0, 1, 0)) }
# mapfn = function(startprice) { return(startprice ^ (1/2)) }
# mapfn = function(startprice) { return(log(startprice)) }
# mapfn = function(startprice) { return(exp(-startprice / 20)) }
# mapfn = function(startprice) { return(scale(log(startprice))) }
# mapfn = function(startprice) { return(sign(sprice.predict.diff) * (abs(sprice.predict.diff) ^ (1/10))) }
# factor
# mapfn = function(PropR) { return(as.factor(ifelse(PropR >= 0.5, "Y", "N"))) }
# mapfn = function(productline, description) { as.factor(gsub(" ", "", productline)) }
# mapfn = function(purpose) { return(relevel(as.factor(purpose), ref="all_other")) }
# mapfn = function(descriptor) { return(plyr::revalue(descriptor, c(
# "ABANDONED BUILDING" = "OTHER",
# "##" = "##"
# ))) }
# mapfn = function(raw) { tfr_raw <- as.character(cut(raw, 5));
# tfr_raw[is.na(tfr_raw)] <- "NA.my";
# return(as.factor(tfr_raw)) }
# mapfn = function(startprice.log10) { return(cut(startprice.log10, 3)) }
# mapfn = function(startprice.log10) { return(cut(sprice.predict.diff, c(-1000, -100, -10, -1, 0, 1, 10, 100, 1000))) }
# , args = c("<arg1>"))
# multiple args
# mapfn = function(PTS, oppPTS) { return(PTS - oppPTS) }
# mapfn = function(startprice.log10.predict, startprice) {
# return(spdiff <- (10 ^ startprice.log10.predict) - startprice) }
# mapfn = function(productline, description) { as.factor(
# paste(gsub(" ", "", productline), as.numeric(nchar(description) > 0), sep = "#")) }
# , args = c("<arg1>", "<arg2>"))
# # If glbObsAll is not sorted in the desired manner
# mapfn=function(Week) { return(coredata(lag(zoo(orderBy(~Week, glbObsAll)$ILI), -2, na.pad=TRUE))) }
# mapfn=function(ILI) { return(coredata(lag(zoo(ILI), -2, na.pad=TRUE))) }
# mapfn=function(ILI.2.lag) { return(log(ILI.2.lag)) }
# glbFeatsDerive[["dummy.my"]] <- list(
# mapfn = function(UniqueID) { return(UniqueID) }
# , args = c("UniqueID"))
glbFeatsDerive[["NDSSName.my"]] <- list(
mapfn = function(NewsDesk, SectionName, SubsectionName) {
descriptor <-
gsub(" ", "", paste(NewsDesk, SectionName, SubsectionName, sep = "#"))
return(plyr::revalue(descriptor, c(NULL
, "#BusinessDay#Dealbook" = "Business#BusinessDay#Dealbook"
, "#BusinessDay#SmallBusiness" = "Business#BusinessDay#SmallBusiness"
, "#Crosswords/Games#" = "Business#Crosswords/Games#"
, "#Open#" = "Business#Technology#"
, "#Technology#" = "Business#Technology#"
, "Business##" = "Business#Technology#"
, "#Arts#" = "Culture#Arts#"
, "Foreign##" = "Foreign#World#"
, "#World#AsiaPacific" = "Foreign#World#AsiaPacific"
, "#N.Y./Region#" = "Metro#N.Y./Region#"
, "#Opinion#" = "OpEd#Opinion#"
, "OpEd##" = "OpEd#Opinion#"
, "#Health#" = "Science#Health#"
, "Science##" = "Science#Health#"
, "Styles#Health#" = "Science#Health#"
, "Styles##" = "Styles##Fashion"
, "Styles#Style#Fashion&Style" = "Styles##Fashion"
, "#Travel#" = "Travel#Travel#"
, "Magazine#Magazine#" = "myOther"
, "National##" = "myOther"
, "National#U.S.#Politics" = "myOther"
, "Sports##" = "myOther"
, "Sports#Sports#" = "myOther"
, "#U.S.#" = "myOther"
)))
}
, args = c("NewsDesk", "SectionName", "SubsectionName"))
# glbFeatsDerive[["<feat.my.sfx>"]] <- list(
# mapfn = function(<arg1>, <arg2>) { return(function(<arg1>, <arg2>)) }
# , args = c("<arg1>", "<arg2>"))
glbFeatsDerive[["WordCount.log1p"]] <- list(
mapfn = function(WordCount) { return(log1p(WordCount)) }
, args = c("WordCount"))
glbFeatsDerive[["WordCount.root2"]] <- list(
mapfn = function(WordCount) { return(WordCount ^ (1/2)) }
, args = c("WordCount"))
glbFeatsDerive[["WordCount.nexp"]] <- list(
mapfn = function(WordCount) { return(exp(-WordCount)) }
, args = c("WordCount"))
#print(summary(glbObsAll$WordCount))
#print(summary(mapfn(glbObsAll$WordCount)))
# glbFeatsDerive[["<var1>"]] <- glbFeatsDerive[["<var2>"]]
glb_derive_vars <- names(glbFeatsDerive)
# tst <- "descr.my"; args_lst <- NULL; for (arg in glbFeatsDerive[[tst]]$args) args_lst[[arg]] <- glbObsAll[, arg]; print(head(args_lst[[arg]])); print(head(drv_vals <- do.call(glbFeatsDerive[[tst]]$mapfn, args_lst)));
# print(which_ix <- which(args_lst[[arg]] == 0.75)); print(drv_vals[which_ix]);
glb_date_vars <- NULL # or c("<date_var>")
glb_date_fmts <- list(); #glb_date_fmts[["<date_var>"]] <- "%m/%e/%y"
glb_date_tzs <- list(); #glb_date_tzs[["<date_var>"]] <- "America/New_York"
#grep("America/New", OlsonNames(), value=TRUE)
glbFeatsPrice <- NULL # or c("<price_var>")
glbFeatsText <- NULL # c("<txt_var>") # NULL #
Sys.setlocale("LC_ALL", "C") # For english## [1] "C/C/C/C/C/en_US.UTF-8"# Text Processing Step: custom modifications not present in txt_munge -> use glbFeatsDerive
# Text Processing Step: universal modifications
glb_txt_munge_filenames_pfx <- "NYTBlogs3_mytxt_"
# Text Processing Step: tolower
# Text Processing Step: myreplacePunctuation
# Text Processing Step: removeWords
glb_txt_stop_words <- list()
# Remember to use unstemmed words
if (!is.null(glbFeatsText)) {
require(tm)
glb_txt_stop_words[["<txt_var>"]] <- sort(c(NULL
# Remove any words from stopwords
# , setdiff(myreplacePunctuation(stopwords("english")), c("<keep_wrd1>", <keep_wrd2>"))
# cor.y.train == NA
# ,unlist(strsplit(paste(c(NULL
# ,"<comma-separated-terms>"
# ), collapse=",")
# freq == 1; keep c("<comma-separated-terms-to-keep>")
# ,<comma-separated-terms>
# chisq.pval high (e.g. == 1); keep c("<comma-separated-terms-to-keep>")
# ,<comma-separated-terms>
# nzv.freqRatio high (e.g. >= glb_nzv_freqCut); keep c("<comma-separated-terms-to-keep>")
# ,<comma-separated-terms>
))
}
#orderBy(~term, glb_post_stem_words_terms_df_lst[[txt_var]][grep("^2", glb_post_stem_words_terms_df_lst[[txt_var]]$term), ])
#glbObsAll[glb_post_stem_words_terms_mtrx_lst[[txt_var]][, 6] > 0, glbFeatsText]
# To identify terms with a specific freq
#paste0(sort(subset(glb_post_stop_words_terms_df_lst[[txt_var]], freq == 1)$term), collapse = ",")
#paste0(sort(subset(glb_post_stem_words_terms_df_lst[[txt_var]], freq <= 2)$term), collapse = ",")
# To identify terms with a specific freq &
# are not stemmed together later OR is value of color.fctr (e.g. gold)
#paste0(sort(subset(glb_post_stop_words_terms_df_lst[[txt_var]], (freq == 1) & !(term %in% c("blacked","blemish","blocked","blocks","buying","cables","careful","carefully","changed","changing","chargers","cleanly","cleared","connect","connects","connected","contains","cosmetics","default","defaulting","defective","definitely","describe","described","devices","displays","drop","drops","engravement","excellant","excellently","feels","fix","flawlessly","frame","framing","gentle","gold","guarantee","guarantees","handled","handling","having","install","iphone","iphones","keeped","keeps","known","lights","line","lining","liquid","liquidation","looking","lots","manuals","manufacture","minis","most","mostly","network","networks","noted","opening","operated","performance","performs","person","personalized","photograph","physically","placed","places","powering","pre","previously","products","protection","purchasing","returned","rotate","rotation","running","sales","second","seconds","shipped","shuts","sides","skin","skinned","sticker","storing","thats","theres","touching","unusable","update","updates","upgrade","weeks","wrapped","verified","verify") ))$term), collapse = ",")
#print(subset(glb_post_stem_words_terms_df_lst[[txt_var]], (freq <= 2)))
#glbObsAll[which(terms_mtrx[, 229] > 0), glbFeatsText]
# To identify terms with cor.y == NA
#orderBy(~-freq+term, subset(glb_post_stop_words_terms_df_lst[[txt_var]], is.na(cor.y)))
#paste(sort(subset(glb_post_stop_words_terms_df_lst[[txt_var]], is.na(cor.y))[, "term"]), collapse=",")
#orderBy(~-freq+term, subset(glb_post_stem_words_terms_df_lst[[txt_var]], is.na(cor.y)))
# To identify terms with low cor.y.abs
#head(orderBy(~cor.y.abs+freq+term, subset(glb_post_stem_words_terms_df_lst[[txt_var]], !is.na(cor.y))), 5)
# To identify terms with high chisq.pval
#subset(glb_post_stem_words_terms_df_lst[[txt_var]], chisq.pval > 0.99)
#paste0(sort(subset(glb_post_stem_words_terms_df_lst[[txt_var]], (chisq.pval > 0.99) & (freq <= 10))$term), collapse=",")
#paste0(sort(subset(glb_post_stem_words_terms_df_lst[[txt_var]], (chisq.pval > 0.9))$term), collapse=",")
#head(orderBy(~-chisq.pval+freq+term, glb_post_stem_words_terms_df_lst[[txt_var]]), 5)
#glbObsAll[glb_post_stem_words_terms_mtrx_lst[[txt_var]][, 68] > 0, glbFeatsText]
#orderBy(~term, glb_post_stem_words_terms_df_lst[[txt_var]][grep("^m", glb_post_stem_words_terms_df_lst[[txt_var]]$term), ])
# To identify terms with high nzv.freqRatio
#summary(glb_post_stem_words_terms_df_lst[[txt_var]]$nzv.freqRatio)
#paste0(sort(setdiff(subset(glb_post_stem_words_terms_df_lst[[txt_var]], (nzv.freqRatio >= glb_nzv_freqCut) & (freq < 10) & (chisq.pval >= 0.05))$term, c( "128gb","3g","4g","gold","ipad1","ipad3","ipad4","ipadair2","ipadmini2","manufactur","spacegray","sprint","tmobil","verizon","wifion"))), collapse=",")
# To identify obs with a txt term
#tail(orderBy(~-freq+term, glb_post_stop_words_terms_df_lst[[txt_var]]), 20)
#mydspObs(list(descr.my.contains="non"), cols=c("color", "carrier", "cellular", "storage"))
#grep("ever", dimnames(terms_stop_mtrx)$Terms)
#which(terms_stop_mtrx[, grep("ipad", dimnames(terms_stop_mtrx)$Terms)] > 0)
#glbObsAll[which(terms_stop_mtrx[, grep("16", dimnames(terms_stop_mtrx)$Terms)[1]] > 0), c(glb_category_var, "storage", txt_var)]
# To identify whether terms shd be synonyms
#orderBy(~term, glb_post_stop_words_terms_df_lst[[txt_var]][grep("^moder", glb_post_stop_words_terms_df_lst[[txt_var]]$term), ])
# term_row_df <- glb_post_stop_words_terms_df_lst[[txt_var]][grep("^came$", glb_post_stop_words_terms_df_lst[[txt_var]]$term), ]
#
# cor(glb_post_stop_words_terms_mtrx_lst[[txt_var]][glbObsAll$.lcn == "Fit", term_row_df$pos], glbObsTrn[, glb_rsp_var], use="pairwise.complete.obs")
# To identify which stopped words are "close" to a txt term
#sort(cluster_vars)
# Text Processing Step: stemDocument
# To identify stemmed txt terms
#glb_post_stop_words_terms_df_lst[[txt_var]][grep("condit", glb_post_stop_words_terms_df_lst[[txt_var]]$term), ]
#orderBy(~term, glb_post_stem_words_terms_df_lst[[txt_var]][grep("^con", glb_post_stem_words_terms_df_lst[[txt_var]]$term), ])
#glbObsAll[which(terms_stem_mtrx[, grep("use", dimnames(terms_stem_mtrx)$Terms)[[1]]] > 0), c(glb_id_var, "productline", txt_var)]
#glbObsAll[which(TfIdf_stem_mtrx[, 191] > 0), c(glb_id_var, glb_category_var, txt_var)]
#which(glbObsAll$UniqueID %in% c(11915, 11926, 12198))
# Text Processing Step: mycombineSynonyms
# To identify which terms are associated with not -> combine "could not" & "couldn't"
#findAssocs(glb_full_DTM_lst[[txt_var]], "not", 0.05)
# To identify which synonyms should be combined
#orderBy(~term, glb_post_stem_words_terms_df_lst[[txt_var]][grep("^c", glb_post_stem_words_terms_df_lst[[txt_var]]$term), ])
chk_comb_cor <- function(syn_lst) {
# cor(terms_stem_mtrx[glbObsAll$.src == "Train", grep("^(damag|dent|ding)$", dimnames(terms_stem_mtrx)[[2]])], glbObsTrn[, glb_rsp_var], use="pairwise.complete.obs")
print(subset(glb_post_stem_words_terms_df_lst[[txt_var]], term %in% syn_lst$syns))
print(subset(get_corpus_terms(tm_map(glb_txt_corpus_lst[[txt_var]], mycombineSynonyms, list(syn_lst), lazy=FALSE)), term == syn_lst$word))
# cor(terms_stop_mtrx[glbObsAll$.src == "Train", grep("^(damage|dent|ding)$", dimnames(terms_stop_mtrx)[[2]])], glbObsTrn[, glb_rsp_var], use="pairwise.complete.obs")
# cor(rowSums(terms_stop_mtrx[glbObsAll$.src == "Train", grep("^(damage|dent|ding)$", dimnames(terms_stop_mtrx)[[2]])]), glbObsTrn[, glb_rsp_var], use="pairwise.complete.obs")
}
#chk_comb_cor(syn_lst=list(word="cabl", syns=c("cabl", "cord")))
#chk_comb_cor(syn_lst=list(word="damag", syns=c("damag", "dent", "ding")))
#chk_comb_cor(syn_lst=list(word="dent", syns=c("dent", "ding")))
#chk_comb_cor(syn_lst=list(word="use", syns=c("use", "usag")))
glb_txt_synonyms <- list()
#glb_txt_synonyms[["<txt_var>"]] <- list(NULL
# , list(word="<stem1>", syns=c("<stem1>", "<stem1_2>"))
# )
# options include: "weightTf", "myweightTflog1p", "myweightTfsqrt", "weightTfIdf", "weightBM25"
glb_txt_terms_control <- list(weighting = "weightTfIdf" # : default
# termFreq selection criteria across obs: tm default: list(global=c(1, Inf))
, bounds = list(global = c(1, Inf))
# wordLengths selection criteria: tm default: c(3, Inf)
, wordLengths = c(1, Inf)
)
glb_txt_cor_var <- glb_rsp_var # : default # or c(<feat>)
# select one from c("union.top.val.cor", "top.cor", "top.val", default: "top.chisq", "sparse")
glbFeatsTextFilter <- "top.chisq"
glbFeatsTextTermsMax <- rep(10, length(glbFeatsText)) # :default
names(glbFeatsTextTermsMax) <- glbFeatsText
# Text Processing Step: extractAssoc
glbFeatsTextAssocCor <- rep(1, length(glbFeatsText)) # :default
names(glbFeatsTextAssocCor) <- glbFeatsText
# Remember to use stemmed terms
glb_important_terms <- list()
# Text Processing Step: extractPatterns (ngrams)
glbFeatsTextPatterns <- list()
#glbFeatsTextPatterns[[<txt_var>>]] <- list()
#glbFeatsTextPatterns[[<txt_var>>]] <- c(metropolitan.diary.colon = "Metropolitan Diary:")
# Have to set it even if it is not used
# Properties:
# numrows(glb_feats_df) << numrows(glbObsFit
# Select terms that appear in at least 0.2 * O(FP/FN(glbObsOOB)) ???
# numrows(glbObsOOB) = 1.1 * numrows(glbObsNew) ???
glb_sprs_thresholds <- NULL # or c(<txt_var1> = 0.988, <txt_var2> = 0.970, <txt_var3> = 0.970)
glbFctrMaxUniqVals <- 21 # default: 20
glb_impute_na_data <- FALSE # or TRUE
glb_mice_complete.seed <- 144 # or any integer
glb_cluster <- FALSE # : default or TRUE
glb_cluster.seed <- 189 # or any integer
glb_cluster_entropy_var <- glb_rsp_var # c(glb_rsp_var, as.factor(cut(glb_rsp_var, 3)), default: NULL)
glbFeatsTextClusterVarsExclude <- FALSE # default FALSE
glb_interaction_only_feats <- NULL # : default or c(<parent_feat> = "<child_feat>")
glb_nzv_freqCut <- 19 # 19 : caret default
glb_nzv_uniqueCut <- 10 # 10 : caret default
glbRFESizes <- list()
#glbRFESizes[["mdlFamily"]] <- c(4, 8, 16, 32, 64, 67, 68, 69) # Accuracy@69/70 = 0.8258
glbObsFitOutliers <- list()
# If outliers.n >= 10; consider concatenation of interaction vars
# glbObsFitOutliers[["<mdlFamily>"]] <- c(NULL
# is.na(.rstudent)
# is.na(.dffits)
# .hatvalues >= 0.99
# -38,167,642 < minmax(.rstudent) < 49,649,823
# , <comma-separated-<glb_id_var>>
# )
glbObsTrnOutliers <- list()
# influence.measures: car::outlier; rstudent; dffits; hatvalues; dfbeta; dfbetas
#mdlId <- "RFE.X.glm"; obs_df <- fitobs_df
#mdlId <- "Final.glm"; obs_df <- trnobs_df
#mdlId <- "CSM2.X.glm"; obs_df <- fitobs_df
#print(outliers <- car::outlierTest(glb_models_lst[[mdlId]]$finalModel))
#mdlIdFamily <- paste0(head(unlist(str_split(mdlId, "\\.")), -1), collapse="."); obs_df <- dplyr::filter_(obs_df, interp(~(!(var %in% glbObsFitOutliers[[mdlIdFamily]])), var = as.name(glb_id_var))); model_diags_df <- cbind(obs_df, data.frame(.rstudent=stats::rstudent(glb_models_lst[[mdlId]]$finalModel)), data.frame(.dffits=stats::dffits(glb_models_lst[[mdlId]]$finalModel)), data.frame(.hatvalues=stats::hatvalues(glb_models_lst[[mdlId]]$finalModel)));print(summary(model_diags_df[, c(".rstudent",".dffits",".hatvalues")])); table(cut(model_diags_df$.hatvalues, breaks=c(0.00, 0.98, 0.99, 1.00)))
#print(subset(model_diags_df, is.na(.rstudent))[, glb_id_var])
#print(subset(model_diags_df, is.na(.dffits))[, glb_id_var])
#print(model_diags_df[which.min(model_diags_df$.dffits), ])
#print(subset(model_diags_df, .hatvalues > 0.99)[, glb_id_var])
#dffits_df <- merge(dffits_df, outliers_df, by="row.names", all.x=TRUE); row.names(dffits_df) <- dffits_df$Row.names; dffits_df <- subset(dffits_df, select=-Row.names)
#dffits_df <- merge(dffits_df, glbObsFit, by="row.names", all.x=TRUE); row.names(dffits_df) <- dffits_df$Row.names; dffits_df <- subset(dffits_df, select=-Row.names)
#subset(dffits_df, !is.na(.Bonf.p))
#mdlId <- "CSM.X.glm"; vars <- myextract_actual_feats(row.names(orderBy(reformulate(c("-", paste0(mdlId, ".importance"))), myget_feats_importance(glb_models_lst[[mdlId]]))));
#model_diags_df <- glb_get_predictions(model_diags_df, mdlId,glb_rsp_var_out)
#obs_ix <- row.names(model_diags_df) %in% names(outliers$rstudent)[1]
#obs_ix <- which(is.na(model_diags_df$.rstudent))
#obs_ix <- which(is.na(model_diags_df$.dffits))
#myplot_parcoord(obs_df=model_diags_df[, c(glb_id_var, glb_category_var, ".rstudent", ".dffits", ".hatvalues", glb_rsp_var, paste0(glb_rsp_var_out, mdlId), vars[1:min(20, length(vars))])], obs_ix=obs_ix, id_var=glb_id_var, category_var=glb_category_var)
#model_diags_df[row.names(model_diags_df) %in% names(outliers$rstudent)[c(1:2)], ]
#ctgry_diags_df <- model_diags_df[model_diags_df[, glb_category_var] %in% c("Unknown#0"), ]
#myplot_parcoord(obs_df=ctgry_diags_df[, c(glb_id_var, glb_category_var, ".rstudent", ".dffits", ".hatvalues", glb_rsp_var, "startprice.log10.predict.RFE.X.glmnet", indep_vars[1:20])], obs_ix=row.names(ctgry_diags_df) %in% names(outliers$rstudent)[1], id_var=glb_id_var, category_var=glb_category_var)
#table(glbObsFit[model_diags_df[, glb_category_var] %in% c("iPad1#1"), "startprice.log10.cut.fctr"])
#glbObsFit[model_diags_df[, glb_category_var] %in% c("iPad1#1"), c(glb_id_var, "startprice")]
# No outliers & .dffits == NaN
#myplot_parcoord(obs_df=model_diags_df[, c(glb_id_var, glb_category_var, glb_rsp_var, "startprice.log10.predict.RFE.X.glmnet", indep_vars[1:10])], obs_ix=seq(1:nrow(model_diags_df))[is.na(model_diags_df$.dffits)], id_var=glb_id_var, category_var=glb_category_var)
# Modify mdlId to (build & extract) "<FamilyId>#<Fit|Trn>#<caretMethod>#<preProc1.preProc2>#<samplingMethod>"
glb_models_lst <- list(); glb_models_df <- data.frame()
# Regression
if (glb_is_regression) {
glbMdlMethods <- c(NULL
# deterministic
#, "lm", # same as glm
, "glm", "bayesglm", "glmnet"
, "rpart"
# non-deterministic
, "gbm", "rf"
# Unknown
, "nnet" , "avNNet" # runs 25 models per cv sample for tunelength=5
, "svmLinear", "svmLinear2"
, "svmPoly" # runs 75 models per cv sample for tunelength=5
, "svmRadial"
, "earth"
, "bagEarth" # Takes a long time
)
} else
# Classification - Add ada (auto feature selection)
if (glb_is_binomial)
glbMdlMethods <- c(NULL
# deterministic
, "bagEarth" # Takes a long time
, "glm", "bayesglm", "glmnet"
, "nnet"
, "rpart"
# non-deterministic
, "gbm"
, "avNNet" # runs 25 models per cv sample for tunelength=5
, "rf"
# Unknown
, "lda", "lda2"
# svm models crash when predict is called -> internal to kernlab it should call predict without .outcome
, "svmLinear", "svmLinear2"
, "svmPoly" # runs 75 models per cv sample for tunelength=5
, "svmRadial"
, "earth"
) else
glbMdlMethods <- c(NULL
# non-deterministic
, "rf"
# Unknown
, "gbm", "rpart"
)
glb_mdl_family_lst <- list(); glb_mdl_feats_lst <- list()
# family: Choose from c("RFE.X", "CSM.X", "All.X", "Best.Interact")
# methods: Choose from c(NULL, <method>, glbMdlMethods)
#glb_mdl_family_lst[["RFE.X"]] <- c("glmnet", "glm") # non-NULL list is mandatory
glb_mdl_family_lst[["All.X"]] <- "glmnet" # non-NULL list is mandatory
#glb_mdl_family_lst[["Best.Interact"]] <- "glmnet" # non-NULL list is mandatory
# Check if interaction features make RFE better
# glb_mdl_family_lst[["CSM.X"]] <- setdiff(glbMdlMethods, c("lda", "lda2")) # crashing due to category:.clusterid ??? #c("glmnet", "glm") # non-NULL list is mandatory
# glb_mdl_feats_lst[["CSM.X"]] <- c(NULL
# , <comma-separated-features-vector>
# )
# dAFeats.CSM.X %<d-% c(NULL
# # Interaction feats up to varImp(RFE.X.glmnet) >= 50
# , <comma-separated-features-vector>
# , setdiff(myextract_actual_feats(predictors(rfe_fit_results)), c(NULL
# , <comma-separated-features-vector>
# ))
# )
# glb_mdl_feats_lst[["CSM.X"]] <- "%<d-% dAFeats.CSM.X"
# Check if tuning parameters make fit better; make it mdlFamily customizable ?
glb_tune_models_df <- data.frame()
#avNNet
# size=[1] 3 5 7 9; decay=[0] 1e-04 0.001 0.01 0.1; bag=[FALSE]; RMSE=1.3300906
#bagEarth
# degree=1 [2] 3; nprune=64 128 256 512 [1024]; RMSE=0.6486663 (up)
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "bagEarth", parameter = "nprune", vals = "256")
# ,data.frame(method = "bagEarth", parameter = "degree", vals = "2")
# ))
#earth
# degree=[1]; nprune=2 [9] 17 25 33; RMSE=0.1334478
#gbm
# shrinkage=0.05 [0.10] 0.15 0.20 0.25; n.trees=100 150 200 [250] 300; interaction.depth=[1] 2 3 4 5; n.minobsinnode=[10]; RMSE=0.2008313
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "gbm", parameter = "shrinkage", min = 0.05, max = 0.25, by = 0.05)
# ,data.frame(method = "gbm", parameter = "n.trees", min = 100, max = 300, by = 50)
# ,data.frame(method = "gbm", parameter = "interaction.depth", min = 1, max = 5, by = 1)
# ,data.frame(method = "gbm", parameter = "n.minobsinnode", min = 10, max = 10, by = 10)
# #seq(from=0.05, to=0.25, by=0.05)
# ))
#glmnet
# alpha=0.100 [0.325] 0.550 0.775 1.000; lambda=0.0005232693 0.0024288010 0.0112734954 [0.0523269304] 0.2428800957; RMSE=0.6164891
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "glmnet", parameter = "alpha", vals = "0.550 0.775 0.8875 0.94375 1.000")
# ,data.frame(method = "glmnet", parameter = "lambda", vals = "9.858855e-05 0.0001971771 0.0009152152 0.0042480525 0.0197177130")
# ))
#nnet
# size=3 5 [7] 9 11; decay=0.0001 0.001 0.01 [0.1] 0.2; RMSE=0.9287422
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "nnet", parameter = "size", vals = "3 5 7 9 11")
# ,data.frame(method = "nnet", parameter = "decay", vals = "0.0001 0.0010 0.0100 0.1000 0.2000")
# ))
#rf # Don't bother; results are not deterministic
# mtry=2 35 68 [101] 134; RMSE=0.1339974
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "rf", parameter = "mtry", vals = "2 5 9 13 17")
# ))
#rpart
# cp=0.020 [0.025] 0.030 0.035 0.040; RMSE=0.1770237
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "rpart", parameter = "cp", vals = "0.004347826 0.008695652 0.017391304 0.021739130 0.034782609")
# ))
#svmLinear
# C=0.01 0.05 [0.10] 0.50 1.00 2.00 3.00 4.00; RMSE=0.1271318; 0.1296718
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "svmLinear", parameter = "C", vals = "0.01 0.05 0.1 0.5 1")
# ))
#svmLinear2
# cost=0.0625 0.1250 [0.25] 0.50 1.00; RMSE=0.1276354
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method = "svmLinear2", parameter = "cost", vals = "0.0625 0.125 0.25 0.5 1")
# ))
#svmPoly
# degree=[1] 2 3 4 5; scale=0.01 0.05 [0.1] 0.5 1; C=0.50 1.00 [2.00] 3.00 4.00; RMSE=0.1276130
# glb_tune_models_df <- myrbind_df(glb_tune_models_df, rbind(data.frame()
# ,data.frame(method="svmPoly", parameter="degree", min=1, max=5, by=1) #seq(1, 5, 1)
# ,data.frame(method="svmPoly", parameter="scale", vals="0.01, 0.05, 0.1, 0.5, 1")
# ,data.frame(method="svmPoly", parameter="C", vals="0.50, 1.00, 2.00, 3.00, 4.00")
# ))
#svmRadial
# sigma=[0.08674323]; C=0.25 0.50 1.00 [2.00] 4.00; RMSE=0.1614957
#glb_to_sav(); all.equal(sav_models_df, glb_models_df)
glb_preproc_methods <- NULL
# c("YeoJohnson", "center.scale", "range", "pca", "ica", "spatialSign")
# Baseline prediction model feature(s)
glb_Baseline_mdl_var <- NULL # or c("<feat>")
glbMdlMetric_terms <- NULL # or matrix(c(
# 0,1,2,3,4,
# 2,0,1,2,3,
# 4,2,0,1,2,
# 6,4,2,0,1,
# 8,6,4,2,0
# ), byrow=TRUE, nrow=5)
glbMdlMetricSummary <- NULL # or "<metric_name>"
glbMdlMetricMaximize <- NULL # or FALSE (TRUE is not the default for both classification & regression)
glbMdlMetricSummaryFn <- NULL # or function(data, lev=NULL, model=NULL) {
# confusion_mtrx <- t(as.matrix(confusionMatrix(data$pred, data$obs)))
# #print(confusion_mtrx)
# #print(confusion_mtrx * glbMdlMetric_terms)
# metric <- sum(confusion_mtrx * glbMdlMetric_terms) / nrow(data)
# names(metric) <- glbMdlMetricSummary
# return(metric)
# }
glb_rcv_n_folds <- 3 # or NULL
glb_rcv_n_repeats <- 3 # or NULL
glb_clf_proba_threshold <- NULL # 0.5
# Model selection criteria
if (glb_is_regression)
glbMdlMetricsEval <- c("min.RMSE.OOB", "max.R.sq.OOB", "max.Adj.R.sq.fit", "min.RMSE.fit")
#glbMdlMetricsEval <- c("min.RMSE.fit", "max.R.sq.fit", "max.Adj.R.sq.fit")
if (glb_is_classification) {
if (glb_is_binomial)
glbMdlMetricsEval <-
c("max.Accuracy.OOB", "max.AUCROCR.OOB", "max.AUCpROC.OOB", "min.aic.fit", "max.Accuracy.fit") else
glbMdlMetricsEval <- c("max.Accuracy.OOB", "max.Kappa.OOB")
}
# select from NULL [no ensemble models], "auto" [all models better than MFO or Baseline], c(mdl_ids in glb_models_lst) [Typically top-rated models in auto]
glb_mdl_ensemble <- NULL
# "%<d-% setdiff(mygetEnsembleAutoMdlIds(), 'CSM.X.rf')"
# c(<comma-separated-mdlIds>
# )
# Only for classifications; for regressions remove "(.*)\\.prob" form the regex
# tmp_fitobs_df <- glbObsFit[, grep(paste0("^", gsub(".", "\\.", glb_rsp_var_out, fixed = TRUE), "CSM\\.X\\.(.*)\\.prob"), names(glbObsFit), value = TRUE)]; cor_mtrx <- cor(tmp_fitobs_df); cor_vctr <- sort(cor_mtrx[row.names(orderBy(~-Overall, varImp(glb_models_lst[["Ensemble.repeatedcv.glmnet"]])$importance))[1], ]); summary(cor_vctr); cor_vctr
#ntv.glm <- glm(reformulate(indep_vars, glb_rsp_var), family = "binomial", data = glbObsFit)
#step.glm <- step(ntv.glm)
glb_sel_mdl_id <- "All.X.glmnet" #select from c(NULL, "All.X.glmnet", "RFE.X.glmnet", <mdlId>)
glb_fin_mdl_id <- NULL #select from c(NULL, glb_sel_mdl_id)
glb_dsp_cols <- c(NULL
# List critical cols excl. glb_id_var, glb_category_var & glb_rsp_var
)
# Output specs
glbOutDataVizFname <- "NYTBlogs3_obsall.csv" # choose from c(NULL, "NYTBlogs3_obsall.csv")
glb_out_obs <- NULL # select from c(NULL : default to "new", "all", "new", "trn")
glb_out_vars_lst <- list()
# glb_id_var will be the first output column, by default
glb_out_vars_lst[["Probability1"]] <-
"%<d-% paste0(glb_rsp_var_out, glb_fin_mdl_id, '.prob')"
# glb_out_vars_lst[[glb_rsp_var_raw]] <- glb_rsp_var_raw
# glb_out_vars_lst[[paste0(head(unlist(strsplit(glb_rsp_var_out, "")), -1), collapse = "")]] <-
glbOutStackFnames <- NULL #: default
# c("ebayipads_txt_assoc1_out_bid1_stack.csv") # manual stack
# c("ebayipads_finmdl_bid1_out_nnet_1.csv") # universal stack
glb_out_pfx <- "NYTBlogs3_base_"
glb_save_envir <- FALSE # or TRUE
# Depict process
glb_analytics_pn <- petrinet(name = "glb_analytics_pn",
trans_df = data.frame(id = 1:6,
name = c("data.training.all","data.new",
"model.selected","model.final",
"data.training.all.prediction","data.new.prediction"),
x=c( -5,-5,-15,-25,-25,-35),
y=c( -5, 5, 0, 0, -5, 5)
),
places_df=data.frame(id=1:4,
name=c("bgn","fit.data.training.all","predict.data.new","end"),
x=c( -0, -20, -30, -40),
y=c( 0, 0, 0, 0),
M0=c( 3, 0, 0, 0)
),
arcs_df=data.frame(
begin=c("bgn","bgn","bgn",
"data.training.all","model.selected","fit.data.training.all",
"fit.data.training.all","model.final",
"data.new","predict.data.new",
"data.training.all.prediction","data.new.prediction"),
end =c("data.training.all","data.new","model.selected",
"fit.data.training.all","fit.data.training.all","model.final",
"data.training.all.prediction","predict.data.new",
"predict.data.new","data.new.prediction",
"end","end")
))
#print(ggplot.petrinet(glb_analytics_pn))
print(ggplot.petrinet(glb_analytics_pn) + coord_flip())## Loading required package: grid
glb_analytics_avl_objs <- NULL
glb_chunks_df <- myadd_chunk(NULL, "import.data")## label step_major step_minor label_minor bgn end elapsed
## 1 import.data 1 0 0 9.037 NA NAStep 1.0: import data
chunk option: eval=
#glb_chunks_df <- myadd_chunk(NULL, "import.data")
glb_to_sav <- function() {
sav_allobs_df <<- glbObsAll
sav_trnobs_df <<- glbObsTrn
if (any(grepl("glbObsFit", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glbObsFit)) sav_fitobs_df <<- glbObsFit
if (any(grepl("glbObsOOB", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glbObsOOB)) sav_OOBobs_df <<- glbObsOOB
if (any(grepl("glbObsNew", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glbObsNew)) {
#print("Attempting to save glbObsNew...")
sav_newobs_df <<- glbObsNew
}
if (any(grepl("glb_ctgry_df", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glb_ctgry_df)) sav_ctgry_df <<- glb_ctgry_df
if (!is.null(glb_models_lst )) sav_models_lst <<- glb_models_lst
if (!is.null(glb_models_df )) sav_models_df <<- glb_models_df
if (any(grepl("glb_feats_df", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glb_feats_df)) sav_feats_df <<- glb_feats_df
if (any(grepl("glb_featsimp_df", ls(envir=globalenv()), fixed=TRUE)) &&
!is.null(glb_featsimp_df)) sav_featsimp_df <<- glb_featsimp_df
}
glbObsTrn <- myimport_data(url=glb_trnng_url, comment="glbObsTrn",
force_header=TRUE)## [1] "Reading file ./data/NYTimesBlogTrain.csv..."
## [1] "dimensions of data in ./data/NYTimesBlogTrain.csv: 6,532 rows x 10 cols"
## NewsDesk SectionName SubsectionName
## 1 Business Crosswords/Games
## 2 Culture Arts
## 3 Business Business Day Dealbook
## 4 Business Business Day Dealbook
## 5 Science Health
## 6 Science Health
## Headline
## 1 More School Daze
## 2 New 96-Page Murakami Work Coming in December
## 3 Public Pension Funds Stay Mum on Corporate Expats
## 4 Boot Camp for Bankers
## 5 Of Little Help to Older Knees
## 6 A Benefit of Legal Marijuana
## Snippet
## 1 A puzzle from Ethan Cooper that reminds me that a bill is due.
## 2 The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His Years of Pilgrimage.
## 3 Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little about the strategy, which could hurt the nations tax base.
## 4 As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service members ideal customers.
## 5 Middle-aged and older patients are unlikely to benefit in the long term from surgery to repair tears in the meniscus, pads of cartilage in the knee, a new review of studies has found.
## 6 A new study has found evidence that legal access to marijuana is associated with fewer opioid overdose deaths, but researchers said their findings should not be used as the basis for the wide adoption of legalized cannabis.
## Abstract
## 1 A puzzle from Ethan Cooper that reminds me that a bill is due.
## 2 The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His Years of Pilgrimage.
## 3 Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little about the strategy, which could hurt the nations tax base.
## 4 As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service members ideal customers.
## 5 Middle-aged and older patients are unlikely to benefit in the long term from surgery to repair tears in the meniscus, pads of cartilage in the knee, a new review of studies has found.
## 6 A new study has found evidence that legal access to marijuana is associated with fewer opioid overdose deaths, but researchers said their findings should not be used as the basis for the wide adoption of legalized cannabis.
## WordCount PubDate Popular UniqueID
## 1 508 2014-09-01 22:00:09 1 1
## 2 285 2014-09-01 21:14:07 0 2
## 3 1211 2014-09-01 21:05:36 0 3
## 4 1405 2014-09-01 20:43:34 1 4
## 5 181 2014-09-01 18:58:51 1 5
## 6 245 2014-09-01 18:52:22 1 6
## NewsDesk SectionName SubsectionName
## 226 Styles
## 995
## 3327
## 4753 Multimedia
## 4802 Business Crosswords/Games
## 6463 TStyle
## Headline
## 226 For Tavi Gevinson, Fashion Takes a Back Seat, for Now
## 995 Reconsidering What to Call an Extremist Group
## 3327 Clinton's Diagnosis of What's Wrong With Politics
## 4753 'Off Color' and on Target About Race in America
## 4802 Daniel Finkel's Circle-Toss Game
## 6463 Entering the Void
## Snippet
## 226 Tavi Gevinson, the teenage fashion star turned Broadway actress, wont be much of a player at New York Fashion Week this season.
## 995 Editors have decided to adjust how The Times refer to an Islamic extremist group that controls territory in Syria and Iraq.
## 3327 Hillary Rodham Clinton continued to laugh off questions about her presidential aspirations on Tuesday, but she did shed some light on what she thinks is wrong in Washington.
## 4753 Off Color, a New York Times video series, looks at how artists of color are making sharp social commentary about race in America through comedy and performance.
## 4802 By math educator Daniel Finkel, a puzzle thats childs play. Can you figure it out?
## 6463 The Spanish artist Miquel Barcel closely examines the basic materials of life in response to Edward Hirsch questioning his own belief in a higher power.
## Abstract
## 226 Tavi Gevinson, the teenage fashion star turned Broadway actress, wont be much of a player at New York Fashion Week this season.
## 995 Editors have decided to adjust how The Times refer to an Islamic extremist group that controls territory in Syria and Iraq.
## 3327 Hillary Rodham Clinton continued to laugh off questions about her presidential aspirations on Tuesday, but she did shed some light on what she thinks is wrong in Washington.
## 4753 Off Color, a New York Times video series, looks at how artists of color are making sharp social commentary about race in America through comedy and performance.
## 4802 By math educator Daniel Finkel, a puzzle thats childs play. Can you figure it out?
## 6463 The Spanish artist Miquel Barcel closely examines the basic materials of life in response to Edward Hirsch questioning his own belief in a higher power.
## WordCount PubDate Popular UniqueID
## 226 459 2014-09-04 16:55:57 0 226
## 995 301 2014-09-15 16:05:13 0 995
## 3327 236 2014-10-14 14:45:51 0 3327
## 4753 393 2014-11-02 05:00:13 0 4753
## 4802 1628 2014-11-03 12:00:04 1 4802
## 6463 264 2014-11-27 12:00:09 0 6463
## NewsDesk SectionName SubsectionName
## 6527 Foreign
## 6528 Opinion Room For Debate
## 6529 Foreign
## 6530 TStyle
## 6531 Multimedia
## 6532 Business
## Headline
## 6527 1914: Russians Dominate in East Poland
## 6528 Finding a Secretary of Defense
## 6529 1889: Metropolitan Opera House Reopens in New York
## 6530 The Daily Gift: Picasso Plates for Creative Dining
## 6531 Racing From New York to Barcelona
## 6532 Math Anxiety: Why Hollywood Makes Robots of Alan Turing and Other Geniuses
## Snippet
## 6527 From the International Herald Tribune archives: Russians dominate in East Poland in 1914.
## 6528 If Chuck Hagel isn't the right Pentagon chief to respond to an onslaught of global crises, who is?
## 6529 From the International Herald Tribune archives: The Metropolitan Opera House reopens in New York in 1889.
## 6530 Each day until Christmas, the editors of T share a new holiday gift idea.
## 6531 A sailboat race from New York to Barcelona was the setting for a thrilling and sometimes terrifying video about this challenging sport.
## 6532 The visionary who stares at formulas written on walls or mirrors or better yet, thin air has become a Hollywood trope. So has the depiction of the genius who cant connect with real people.
## Abstract
## 6527 From the International Herald Tribune archives: Russians dominate in East Poland in 1914.
## 6528 If Chuck Hagel isn't the right Pentagon chief to respond to an onslaught of global crises, who is?
## 6529 From the International Herald Tribune archives: The Metropolitan Opera House reopens in New York in 1889.
## 6530 Each day until Christmas, the editors of T share a new holiday gift idea.
## 6531 A sailboat race from New York to Barcelona was the setting for a thrilling and sometimes terrifying video about this challenging sport.
## 6532 The visionary who stares at formulas written on walls or mirrors or better yet, thin air has become a Hollywood trope. So has the depiction of the genius who cant connect with real people.
## WordCount PubDate Popular UniqueID
## 6527 176 2014-11-30 13:48:40 0 6527
## 6528 1597 2014-11-30 13:27:23 0 6528
## 6529 214 2014-11-30 09:44:57 0 6529
## 6530 61 2014-11-30 09:00:43 0 6530
## 6531 441 2014-11-30 09:00:22 0 6531
## 6532 921 2014-11-30 07:00:40 0 6532
## 'data.frame': 6532 obs. of 10 variables:
## $ NewsDesk : chr "Business" "Culture" "Business" "Business" ...
## $ SectionName : chr "Crosswords/Games" "Arts" "Business Day" "Business Day" ...
## $ SubsectionName: chr "" "" "Dealbook" "Dealbook" ...
## $ Headline : chr "More School Daze" "New 96-Page Murakami Work Coming in December" "Public Pension Funds Stay Mum on Corporate Expats" "Boot Camp for Bankers" ...
## $ Snippet : chr "A puzzle from Ethan Cooper that reminds me that a bill is due." "The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His"| __truncated__ "Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little"| __truncated__ "As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service "| __truncated__ ...
## $ Abstract : chr "A puzzle from Ethan Cooper that reminds me that a bill is due." "The Strange Library will arrive just three and a half months after Mr. Murakamis latest novel, Colorless Tsukuru Tazaki and His"| __truncated__ "Public pension funds have major stakes in American companies moving overseas to cut their tax bills. But they are saying little"| __truncated__ "As they struggle to find new business to bolster sluggish earnings, banks consider the nations 25 million veterans and service "| __truncated__ ...
## $ WordCount : int 508 285 1211 1405 181 245 258 893 1077 188 ...
## $ PubDate : chr "2014-09-01 22:00:09" "2014-09-01 21:14:07" "2014-09-01 21:05:36" "2014-09-01 20:43:34" ...
## $ Popular : int 1 0 0 1 1 1 0 1 1 0 ...
## $ UniqueID : int 1 2 3 4 5 6 7 8 9 10 ...
## - attr(*, "comment")= chr "glbObsTrn"
## NULL# glbObsTrn <- read.delim("data/hygiene.txt", header=TRUE, fill=TRUE, sep="\t",
# fileEncoding='iso-8859-1')
# glbObsTrn <- read.table("data/hygiene.dat.labels", col.names=c("dirty"),
# na.strings="[none]")
# glbObsTrn$review <- readLines("data/hygiene.dat", n =-1)
# comment(glbObsTrn) <- "glbObsTrn"
# glbObsTrn <- data.frame()
# for (symbol in c("Boeing", "CocaCola", "GE", "IBM", "ProcterGamble")) {
# sym_trnobs_df <-
# myimport_data(url=gsub("IBM", symbol, glb_trnng_url), comment="glbObsTrn",
# force_header=TRUE)
# sym_trnobs_df$Symbol <- symbol
# glbObsTrn <- myrbind_df(glbObsTrn, sym_trnobs_df)
# }
# glbObsTrn <-
# glbObsTrn %>% dplyr::filter(Year >= 1999)
if (glb_is_separate_newobs_dataset) {
glbObsNew <- myimport_data(url=glb_newdt_url, comment="glbObsNew",
force_header=TRUE)
# To make plots / stats / checks easier in chunk:inspectORexplore.data
glbObsAll <- myrbind_df(glbObsTrn, glbObsNew);
comment(glbObsAll) <- "glbObsAll"
} else {
glbObsAll <- glbObsTrn; comment(glbObsAll) <- "glbObsAll"
if (!glb_split_entity_newobs_datasets) {
stop("Not implemented yet")
glbObsNew <- glbObsTrn[sample(1:nrow(glbObsTrn),
max(2, nrow(glbObsTrn) / 1000)),]
} else if (glb_split_newdata_method == "condition") {
glbObsNew <- do.call("subset",
list(glbObsTrn, parse(text=glb_split_newdata_condition)))
glbObsTrn <- do.call("subset",
list(glbObsTrn, parse(text=paste0("!(",
glb_split_newdata_condition,
")"))))
} else if (glb_split_newdata_method == "sample") {
require(caTools)
set.seed(glb_split_sample.seed)
split <- sample.split(glbObsTrn[, glb_rsp_var_raw],
SplitRatio=(1-glb_split_newdata_size_ratio))
glbObsNew <- glbObsTrn[!split, ]
glbObsTrn <- glbObsTrn[split ,]
} else if (glb_split_newdata_method == "copy") {
glbObsTrn <- glbObsAll
comment(glbObsTrn) <- "glbObsTrn"
glbObsNew <- glbObsAll
comment(glbObsNew) <- "glbObsNew"
} else stop("glb_split_newdata_method should be %in% c('condition', 'sample', 'copy')")
comment(glbObsNew) <- "glbObsNew"
myprint_df(glbObsNew)
str(glbObsNew)
if (glb_split_entity_newobs_datasets) {
myprint_df(glbObsTrn)
str(glbObsTrn)
}
} ## [1] "Reading file ./data/NYTimesBlogTest.csv..."
## [1] "dimensions of data in ./data/NYTimesBlogTest.csv: 1,870 rows x 9 cols"
## NewsDesk SectionName SubsectionName
## 1 Culture
## 2 Culture Arts
## 3 Business Crosswords/Games
## 4 Business Business Day Dealbook
## 5 Science Health
## 6 Science Health
## Headline
## 1 'Birdman' Tops the Gothams
## 2 'Sleepy Hollow' Recap: A Not-So-Shocking Death
## 3 Drinking Buddy For Falstaff
## 4 Encouraging Public Service, Through Wall Street's 'Revolving Door'
## 5 Therapy Prevents Repeat Suicide Attempts
## 6 Hoping for a Good Death
## Snippet
## 1 The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner.
## 2 In the fall season finale, a question of where the series has many places to go.
## 3 In which Timothy Polin reveals his potty mouth.
## 4 The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than on good public policy.
## 5 Short-term psychotherapy may be an effective way to prevent repeated suicide attempts.
## 6 What I hadnt considered before my fathers heart attack was the precise meaning of not wanting to live hooked up to machines.
## Abstract
## 1 The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner.
## 2 In the fall season finale, a question of where the series has many places to go.
## 3 In which Timothy Polin reveals his potty mouth.
## 4 The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than on good public policy.
## 5 Short-term psychotherapy may be an effective way to prevent repeated suicide attempts.
## 6 What I hadnt considered before my fathers heart attack was the precise meaning of not wanting to live hooked up to machines.
## WordCount PubDate UniqueID
## 1 111 2014-12-01 22:45:24 6533
## 2 558 2014-12-01 22:01:34 6534
## 3 788 2014-12-01 22:00:26 6535
## 4 915 2014-12-01 21:04:13 6536
## 5 213 2014-12-01 19:13:20 6537
## 6 938 2014-12-01 19:05:12 6538
## NewsDesk SectionName SubsectionName
## 3 Business Crosswords/Games
## 334 OpEd Opinion
## 725 TStyle
## 732 Business Business Day Dealbook
## 752 Business Business Day Dealbook
## 864
## Headline
## 3 Drinking Buddy For Falstaff
## 334 Facts & Figures: America’s Unique Take on Maternity Leave
## 725 Ansel Elgort Buttons Up in Brioni
## 732 A Shake-Up as the Financial World Infiltrates Philanthropy
## 752 Coupang, a South Korean E-Commerce Site, Raises $300 Million
## 864 Today in Politics
## Snippet
## 3 In which Timothy Polin reveals his potty mouth.
## 334 In the U.S., paid parental leave is more of a perk than a guarantee.
## 725 The actor brought a tinge of youthfulness to the classic Italian houses retro-tailored look.
## 732 Donor-advised funds help investors get deductions for charitable donations in one year, but society doesnt get the benefit of the money right away.
## 752 The latest financing round underscores Coupangs maturity and its ambitions to one day be a publicly traded company.
## 864 The 113th Congress is concluding with partisan brinksmanship and one last mad scramble for votes to pass a $1.1 trillion spending package.
## Abstract
## 3 In which Timothy Polin reveals his potty mouth.
## 334 In the U.S., paid parental leave is more of a perk than a guarantee.
## 725 The actor brought a tinge of youthfulness to the classic Italian houses retro-tailored look.
## 732 Donor-advised funds help investors get deductions for charitable donations in one year, but society doesnt get the benefit of the money right away.
## 752 The latest financing round underscores Coupangs maturity and its ambitions to one day be a publicly traded company.
## 864 The 113th Congress is concluding with partisan brinksmanship and one last mad scramble for votes to pass a $1.1 trillion spending package.
## WordCount PubDate UniqueID
## 3 788 2014-12-01 22:00:26 6535
## 334 160 2014-12-04 11:45:20 6866
## 725 89 2014-12-10 12:30:47 7257
## 732 1172 2014-12-10 12:00:38 7264
## 752 353 2014-12-10 08:30:41 7284
## 864 1544 2014-12-11 07:09:25 7396
## NewsDesk SectionName SubsectionName
## 1865
## 1866 Business Technology
## 1867 Metro N.Y. / Region
## 1868 Multimedia
## 1869 Foreign World Asia Pacific
## 1870 Science Health
## Headline
## 1865 Today in Politics
## 1866 Uber Suspends Operations in Spain
## 1867 New York Today: The Year in News
## 1868 New Year, Old Memories, in Times Square
## 1869 Hong Kong Police Criticized After 14-Year-Old's Detention
## 1870 The Super-Short Workout and Other Fitness Trends
## Snippet
## 1865 House Republicans are ending the year on a defensive note over Representative Steve Scalises 2002 speech to a white supremacist group.
## 1866 In a first in the growing pushback against Ubers global expansion, a judges ruling barred telecommunications operators and banks from supporting the companys services.
## 1867 Wednesday: The most read stories of 2014, teeth-chattering cold, and its New Years Eve.
## 1868 What happens when you combine Burning Man, Independence Day fireworks, the last day of school and a full-contact Black Friday sale-a-bration? New Years Eve in Times Square.
## 1869 The authorities have been accused of trying to intimidate young pro-democracy protesters and their families after a 14-year-old girl was detained on suspicion of drawing flowers in chalk near government headquarters and sent to a juvenile home.
## 1870 The big story in exercise science this year was the super-short workout, although many other fitness-related themes emerged in 2014.
## Abstract
## 1865 House Republicans are ending the year on a defensive note over Representative Steve Scalises 2002 speech to a white supremacist group.
## 1866 In a first in the growing pushback against Ubers global expansion, a judges ruling barred telecommunications operators and banks from supporting the companys services.
## 1867 Wednesday: The most read stories of 2014, teeth-chattering cold, and its New Years Eve.
## 1868 What happens when you combine Burning Man, Independence Day fireworks, the last day of school and a full-contact Black Friday sale-a-bration? New Years Eve in Times Square.
## 1869 The authorities have been accused of trying to intimidate young pro-democracy protesters and their families after a 14-year-old girl was detained on suspicion of drawing flowers in chalk near government headquarters and sent to a juvenile home.
## 1870 The big story in exercise science this year was the super-short workout, although many other fitness-related themes emerged in 2014.
## WordCount PubDate UniqueID
## 1865 1616 2014-12-31 07:03:46 8397
## 1866 292 2014-12-31 06:09:32 8398
## 1867 1010 2014-12-31 06:06:58 8399
## 1868 387 2014-12-31 05:00:19 8400
## 1869 717 2014-12-31 04:16:29 8401
## 1870 818 2014-12-31 00:01:10 8402
## 'data.frame': 1870 obs. of 9 variables:
## $ NewsDesk : chr "Culture" "Culture" "Business" "Business" ...
## $ SectionName : chr "" "Arts" "Crosswords/Games" "Business Day" ...
## $ SubsectionName: chr "" "" "" "Dealbook" ...
## $ Headline : chr "'Birdman' Tops the Gothams" "'Sleepy Hollow' Recap: A Not-So-Shocking Death" "Drinking Buddy For Falstaff" "Encouraging Public Service, Through Wall Street's 'Revolving Door'" ...
## $ Snippet : chr "The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner." "In the fall season finale, a question of where the series has many places to go." "In which Timothy Polin reveals his potty mouth." "The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than "| __truncated__ ...
## $ Abstract : chr "The backstage tale won two awards; Citizenfour, the Edward Snowden documentary, was also a winner." "In the fall season finale, a question of where the series has many places to go." "In which Timothy Polin reveals his potty mouth." "The debate about pay for Wall Street executives who take government jobs appears to be based more on a populist shakedown than "| __truncated__ ...
## $ WordCount : int 111 558 788 915 213 938 1336 2644 752 99 ...
## $ PubDate : chr "2014-12-01 22:45:24" "2014-12-01 22:01:34" "2014-12-01 22:00:26" "2014-12-01 21:04:13" ...
## $ UniqueID : int 6533 6534 6535 6536 6537 6538 6539 6540 6541 6542 ...
## - attr(*, "comment")= chr "glbObsNew"
## NULLif ((num_nas <- sum(is.na(glbObsTrn[, glb_rsp_var_raw]))) > 0)
stop("glbObsTrn$", glb_rsp_var_raw, " contains NAs for ", num_nas, " obs")
if (nrow(glbObsTrn) == nrow(glbObsAll))
warning("glbObsTrn same as glbObsAll")
if (nrow(glbObsNew) == nrow(glbObsAll))
warning("glbObsNew same as glbObsAll")
if (length(glb_drop_vars) > 0) {
warning("dropping vars: ", paste0(glb_drop_vars, collapse=", "))
glbObsAll <- glbObsAll[, setdiff(names(glbObsAll), glb_drop_vars)]
glbObsTrn <- glbObsTrn[, setdiff(names(glbObsTrn), glb_drop_vars)]
glbObsNew <- glbObsNew[, setdiff(names(glbObsNew), glb_drop_vars)]
}
#stop(here"); sav_allobs_df <- glbObsAll # glbObsAll <- sav_allobs_df
# Combine trnent & newobs into glbObsAll for easier manipulation
glbObsTrn$.src <- "Train"; glbObsNew$.src <- "Test";
glbFeatsExclude <- union(glbFeatsExclude, ".src")
glbObsAll <- myrbind_df(glbObsTrn, glbObsNew)
comment(glbObsAll) <- "glbObsAll"
# Check for duplicates in glb_id_var
if (length(glb_id_var) == 0) {
warning("using .rownames as identifiers for observations")
glbObsAll$.rownames <- rownames(glbObsAll)
glbObsTrn$.rownames <- rownames(subset(glbObsAll, .src == "Train"))
glbObsNew$.rownames <- rownames(subset(glbObsAll, .src == "Test"))
glb_id_var <- ".rownames"
}
if (sum(duplicated(glbObsAll[, glb_id_var, FALSE])) > 0)
stop(glb_id_var, " duplicated in glbObsAll")
glbFeatsExclude <- union(glbFeatsExclude, glb_id_var)
glbObsAll <- orderBy(reformulate(glb_id_var), glbObsAll)
glbObsTrn <- glbObsNew <- NULL
# For Tableau
if (!is.null(glbOutDataVizFname))
write.csv(glbObsAll, glbOutDataVizFname, row.names=FALSE)
# - Merge glb_obs_stack_condition & glbObsDropCondition
# - Derive glb_obs_stack|drop_chk_vars from condition automatically
# - Implement glb_obs_stack_condition & glb_obs_stack_chk_vars options
dsp_partition_stats <- function(obs_df, vars=NULL) {
lcl_vars <- NULL
for (var in c(vars, glb_rsp_var_raw)) {
if ((length(unique(obs_df[, var])) > 5) && is.numeric(obs_df[, var])) {
cut_var <- paste0(var, ".cut.fctr")
obs_df[, cut_var] <- cut(obs_df[, var], 3)
lcl_vars <- union(lcl_vars, cut_var)
} else lcl_vars <- union(lcl_vars, var)
}
print("Partition stats:")
print(mycreate_sqlxtab_df(obs_df, union(lcl_vars, ".src")))
for (var in lcl_vars) {
print(freq_df <- mycreate_sqlxtab_df(obs_df, union(var, ".src")))
print(myplot_hbar(freq_df, ".src", ".n", colorcol_name=var))
}
print(mycreate_sqlxtab_df(obs_df, ".src"))
}
myget_symbols <- function(txt) {
if (is.null(txt)) return(NULL)
#print(getParseData(parse(text=txt, keep.source=TRUE)))
return(unique(subset(getParseData(parse(text=txt, keep.source=TRUE)),
token == "SYMBOL")$text))
}
# tokens <- unlist(strsplit(gsub("[[:punct:]|[:space:]]", " ", glbObsDropCondition), " "))
# tokens <- tokens[tokens != ""]
# glb_obs_drop_chk_vars <- c("biddable") # or NULL
dsp_partition_stats(obs_df=glbObsAll, vars=myget_symbols(glbObsDropCondition))## [1] "Partition stats:"## Loading required package: sqldf
## Loading required package: gsubfn
## Loading required package: proto
## Loading required package: RSQLite
## Loading required package: DBI
## Loading required package: tcltk## Popular .src .n
## 1 0 Train 5439
## 2 NA Test 1870
## 3 1 Train 1093
## Popular .src .n
## 1 0 Train 5439
## 2 NA Test 1870
## 3 1 Train 1093
## .src .n
## 1 Train 6532
## 2 Test 1870if (!is.null(glbObsDropCondition)) {
print(sprintf("Running glbObsDropCondition filter: %s", glbObsDropCondition))
glbObsAll <- do.call("subset",
list(glbObsAll, parse(text=paste0("!(", glbObsDropCondition, ")"))))
dsp_partition_stats(obs_df=glbObsAll, vars=myget_symbols(glbObsDropCondition))
}
# Check for duplicates by all features
# Refactor to utilize glbSpecs
require(dplyr)## Loading required package: dplyr
##
## Attaching package: 'dplyr'
##
## The following objects are masked from 'package:stats':
##
## filter, lag
##
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, unionrequire(lazyeval)## Loading required package: lazyevalrequire(gdata)## Loading required package: gdata
## gdata: read.xls support for 'XLS' (Excel 97-2004) files ENABLED.
##
## gdata: read.xls support for 'XLSX' (Excel 2007+) files ENABLED.
##
## Attaching package: 'gdata'
##
## The following objects are masked from 'package:dplyr':
##
## combine, first, last
##
## The following object is masked from 'package:stats':
##
## nobs
##
## The following object is masked from 'package:utils':
##
## object.size#print(names(glbObsAll))
dupObsAll <- glbObsAll[duplicated2(subset(glbObsAll,
select = -c(UniqueID, Popular, .src))), ]
dupObsAllIx <- glbObsAll %>%
dplyr::select_(.dots = setdiff(names(glbObsAll),
c(glb_id_var, glb_rsp_var_raw, ".src"))) %>%
#dplyr::select(-c(UniqueID, Popular, .src)) %>%
duplicated2()
dupObsAll <- glbObsAll[dupObsAllIx, ] %>%
dplyr::arrange_(.dots = setdiff(names(glbObsAll),
c(glb_id_var, glb_rsp_var_raw, ".src")))
print(sprintf("Found %d duplicates by all features:", nrow(dupObsAll)))## [1] "Found 0 duplicates by all features:"myprint_df(dupObsAll)## [1] NewsDesk SectionName SubsectionName Headline
## [5] Snippet Abstract WordCount PubDate
## [9] Popular UniqueID .src
## <0 rows> (or 0-length row.names)# print(dupObsAll[, c(glb_id_var, glb_rsp_var_raw,
# "description", "startprice", "biddable")])
# write.csv(dupObsAll[, c("UniqueID"), FALSE], "ebayipads_dups.csv", row.names=FALSE)
if (nrow(dupObsAll) > 0) {
dupobs_df <- tidyr::unite(dupObsAll, "allfeats", -c(sold, UniqueID, .src), sep="#")
# dupobs_df <- dplyr::group_by(dupobs_df, allfeats)
# dupobs_df <- dupobs_df[, "UniqueID", FALSE]
# dupobs_df <- ungroup(dupobs_df)
#
# dupobs_df$.rownames <- row.names(dupobs_df)
grpobs_df <- data.frame(allfeats=unique(dupobs_df[, "allfeats"]))
grpobs_df$.grpid <- row.names(grpobs_df)
dupobs_df <- merge(dupobs_df, grpobs_df)
# dupobs_tbl <- table(dupobs_df$.grpid)
# print(max(dupobs_tbl))
# print(dupobs_tbl[which.max(dupobs_tbl)])
# print(dupobs_df[dupobs_df$.grpid == names(dupobs_tbl[which.max(dupobs_tbl)]), ])
# print(dupobs_df[dupobs_df$.grpid == 106, ])
# for (grpid in c(9, 17, 31, 36, 53))
# print(dupobs_df[dupobs_df$.grpid == grpid, ])
dupgrps_df <- as.data.frame(table(dupobs_df$.grpid, dupobs_df$sold, useNA="ifany"))
names(dupgrps_df)[c(1,2)] <- c(".grpid", "sold")
dupgrps_df$.grpid <- as.numeric(as.character(dupgrps_df$.grpid))
dupgrps_df <- tidyr::spread(dupgrps_df, sold, Freq)
names(dupgrps_df)[-1] <- paste("sold", names(dupgrps_df)[-1], sep=".")
dupgrps_df$.freq <- sapply(1:nrow(dupgrps_df), function(row) sum(dupgrps_df[row, -1]))
myprint_df(orderBy(~-.freq, dupgrps_df))
print("sold Conflicts:")
print(subset(dupgrps_df, (sold.0 > 0) & (sold.1 > 0)))
#dupobs_df[dupobs_df$.grpid == 4, ]
glbObsAll <- merge(glbObsAll, dupobs_df[, c(glb_id_var, ".grpid")],
by=glb_id_var, all.x=TRUE)
if (nrow(subset(dupgrps_df, (sold.0 > 0) & (sold.1 > 0) & (sold.0 != sold.1))) > 0)
stop("Duplicate conflicts are resolvable")
#subset(glbObsAll, .grpid %in% c(25))
#mydspObs(list(productline.contains="iPad 1", storage.contains="16", color.contains="Black", carrier.contains="None", cellular.contains="0", condition.contains="Used", startprice=80), cols=c("productline", "storage", "color", "carrier", "cellular", "condition", "startprice", "sold"))
print("Test & Train Groups:")
print(subset(dupgrps_df, (sold.NA > 0)))
glbFeatsExclude <- c(".grpid", glbFeatsExclude)
}
if (!is.null(glbInpMerge)) {
print("Running glbInpMerge specs...")
obsMrg <- data.frame()
for (fName in glbInpMerge$fnames) {
print(sprintf(" Appending rows from %s...", fName))
obsMrg <- rbind(obsMrg, read.csv(fName))
}
glbObsAll <- merge(glbObsAll, obsMrg, all.x = TRUE)
}
dsp_partition_stats(obs_df = glbObsAll,
vars = myget_symbols(glb_obs_repartition_train_condition))## [1] "Partition stats:"
## Popular .src .n
## 1 0 Train 5439
## 2 NA Test 1870
## 3 1 Train 1093
## Popular .src .n
## 1 0 Train 5439
## 2 NA Test 1870
## 3 1 Train 1093
## .src .n
## 1 Train 6532
## 2 Test 1870if (!is.null(glb_obs_repartition_train_condition)) {
print(sprintf("Running glb_obs_repartition_train_condition filter: %s",
glb_obs_repartition_train_condition))
# glbObsAll <- mutate(glbObsAll, .src=ifelse(!is.na(sold) & (sold == 1),
# "Train", "Test"))
# glbObsAll <- mutate_(glbObsAll,
# .src=interp(ifelse(eval(parse(text="!is.na(sold) & (sold == 1)")),
# "Train", "Test")))
# glbObsAll <- within(glbObsAll, {
# .src <- ifelse(eval(parse(text="!is.na(sold) & (sold == 1)")),
# "Train", "Test")
# })
# glbObsAll <- within(glbObsAll, {
# if(eval(parse(text="!is.na(sold) & (sold == 1)"))) .src <- "Train" else
# .src <- "Test"
# })
# with(glbObsAll, {
# src <- ifelse(eval(parse(text="!is.na(sold) & (sold == 1)")),
# "Train", "Test")
# })
# glbObsAll$.src <- sapply(1:nrow(glbObsAll), function (row_ix) ifelse)
# glbObsAll[parse(text=paste0("!(", glbObsDropCondition, ")")), ".src"] <- do.call("subset",
# list(glbObsAll, ))
glbObsTrn <- do.call("subset", list(glbObsAll,
parse(text=paste0(" (", glb_obs_repartition_train_condition, ")"))))
glbObsTrn$.src <- "Train"
glbObsNew <- do.call("subset", list(glbObsAll,
parse(text=paste0("!(", glb_obs_repartition_train_condition, ")"))))
glbObsNew$.src <- "Test"
glbObsAll <- rbind(glbObsTrn, glbObsNew)
dsp_partition_stats(obs_df = glbObsAll,
vars = myget_symbols(glb_obs_repartition_train_condition))
}
glb_chunks_df <- myadd_chunk(glb_chunks_df, "inspect.data", major.inc=TRUE)## label step_major step_minor label_minor bgn end elapsed
## 1 import.data 1 0 0 9.037 18.458 9.421
## 2 inspect.data 2 0 0 18.458 NA NAStep 2.0: inspect data
#print(str(glbObsAll))
#View(glbObsAll)
dsp_class_dstrb <- function(var) {
xtab_df <- mycreate_xtab_df(glbObsAll, c(".src", var))
rownames(xtab_df) <- xtab_df$.src
xtab_df <- subset(xtab_df, select=-.src)
print(xtab_df)
print(xtab_df / rowSums(xtab_df, na.rm=TRUE))
}
# Performed repeatedly in other chunks
glb_chk_data <- function(featsExclude = glbFeatsExclude,
fctrMaxUniqVals = glbFctrMaxUniqVals) {
# Histogram of predictor in glbObsTrn & glbObsNew
print(myplot_histogram(glbObsAll, glb_rsp_var_raw) + facet_wrap(~ .src))
if (glb_is_classification)
dsp_class_dstrb(var=ifelse(glb_rsp_var %in% names(glbObsAll),
glb_rsp_var, glb_rsp_var_raw))
mycheck_problem_data(glbObsAll, featsExclude, fctrMaxUniqVals)
}
glb_chk_data()## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.## Warning: Removed 1870 rows containing non-finite values (stat_bin).## Loading required package: reshape2
## Popular.0 Popular.1 Popular.NA
## Test NA NA 1870
## Train 5439 1093 NA
## Popular.0 Popular.1 Popular.NA
## Test NA NA 1
## Train 0.8326699 0.1673301 NA
## [1] "numeric data missing in glbObsAll: "
## Popular
## 1870
## [1] "numeric data w/ 0s in glbObsAll: "
## WordCount Popular
## 109 5439
## [1] "numeric data w/ Infs in glbObsAll: "
## named integer(0)
## [1] "numeric data w/ NaNs in glbObsAll: "
## named integer(0)
## [1] "string data missing in glbObsAll: "
## NewsDesk SectionName SubsectionName Headline Snippet
## 2408 2899 6176 0 13
## Abstract PubDate
## 17 0# Create new features that help diagnostics
if (!is.null(glb_map_rsp_raw_to_var)) {
glbObsAll[, glb_rsp_var] <-
glb_map_rsp_raw_to_var(glbObsAll[, glb_rsp_var_raw])
mycheck_map_results(mapd_df=glbObsAll,
from_col_name=glb_rsp_var_raw, to_col_name=glb_rsp_var)
if (glb_is_classification) dsp_class_dstrb(glb_rsp_var)
}## Popular Popular.fctr .n
## 1 0 N 5439
## 2 NA <NA> 1870
## 3 1 Y 1093## Warning: Removed 1 rows containing missing values (position_stack).
## Popular.fctr.N Popular.fctr.Y Popular.fctr.NA
## Test NA NA 1870
## Train 5439 1093 NA
## Popular.fctr.N Popular.fctr.Y Popular.fctr.NA
## Test NA NA 1
## Train 0.8326699 0.1673301 NA# check distribution of all numeric data
dsp_numeric_feats_dstrb <- function(feats_vctr) {
grid.newpage()
pushViewport(viewport(layout = grid.layout(ceiling(length(feats_vctr) / 2.0), 2)))
pltIx <- 1
for (feat in feats_vctr) {
#print(sprintf("feat: %s", feat))
if (glb_is_regression)
gp <- myplot_scatter(df=glbObsAll, ycol_name=glb_rsp_var, xcol_name=feat,
smooth=TRUE)
if (glb_is_classification)
#gp <- myplot_box(df=glbObsAll, ycol_names=feat, xcol_name=glb_rsp_var)
gp <- myplot_violin(glbObsAll, ycol_names = feat, xcol_name = glb_rsp_var)
if (inherits(glbObsAll[, feat], "factor"))
gp <- gp + facet_wrap(reformulate(feat))
print(gp + labs(title = feat),
vp = viewport(layout.pos.row = ceiling(pltIx / 2.0),
layout.pos.col = ((pltIx - 1) %% 2) + 1))
pltIx <- pltIx + 1
}
}
# dsp_numeric_feats_dstrb(setdiff(names(glbObsAll), union(myfind_chr_cols_df(glbObsAll), c(glb_rsp_var_raw, glb_rsp_var)))[2:6]) # dsp_numeric_feats_dstrb(c("startprice", "sprice.root3", "sprice.predict.diff"))
add_new_diag_feats <- function(obs_df, ref_df=glbObsAll) {
require(plyr)
set.seed(169)
obs_df <- mutate(obs_df,
# <col_name>.NA=is.na(<col_name>),
# <col_name>.fctr=factor(<col_name>,
# as.factor(union(obs_df$<col_name>, obs_twin_df$<col_name>))),
# <col_name>.fctr=relevel(factor(<col_name>,
# as.factor(union(obs_df$<col_name>, obs_twin_df$<col_name>))),
# "<ref_val>"),
# <col2_name>.fctr=relevel(factor(ifelse(<col1_name> == <val>, "<oth_val>", "<ref_val>")),
# as.factor(c("R", "<ref_val>")),
# ref="<ref_val>"),
# This doesn't work - use sapply instead
# <col_name>.fctr_num=grep(<col_name>, levels(<col_name>.fctr)),
#
# Date.my=as.Date(strptime(Date, "%m/%d/%y %H:%M")),
# Year=year(Date.my),
# Month=months(Date.my),
# Weekday=weekdays(Date.my)
# <col_name>=<table>[as.character(<col2_name>)],
# <col_name>=as.numeric(<col2_name>),
# <col_name> = trunc(<col2_name> / 100),
.rnorm = rnorm(n=nrow(obs_df))
)
# If levels of a factor are different across obs_df & glbObsNew; predict.glm fails
# Transformations not handled by mutate
# obs_df$<col_name>.fctr.num <- sapply(1:nrow(obs_df),
# function(row_ix) grep(obs_df[row_ix, "<col_name>"],
# levels(obs_df[row_ix, "<col_name>.fctr"])))
#print(summary(obs_df))
#print(sapply(names(obs_df), function(col) sum(is.na(obs_df[, col]))))
return(obs_df)
}
glbObsAll <- add_new_diag_feats(glbObsAll)## Loading required package: plyr
## -------------------------------------------------------------------------
## You have loaded plyr after dplyr - this is likely to cause problems.
## If you need functions from both plyr and dplyr, please load plyr first, then dplyr:
## library(plyr); library(dplyr)
## -------------------------------------------------------------------------
##
## Attaching package: 'plyr'
##
## The following objects are masked from 'package:dplyr':
##
## arrange, count, desc, failwith, id, mutate, rename, summarise,
## summarizerequire(dplyr)
#stop(here"); sav_allobs_df <- glbObsAll # glbObsAll <- sav_allobs_df
# Merge some <descriptor>
# glbObsAll$<descriptor>.my <- glbObsAll$<descriptor>
# glbObsAll[grepl("\\bAIRPORT\\b", glbObsAll$<descriptor>.my),
# "<descriptor>.my"] <- "AIRPORT"
# Check distributions of newly transformed / extracted vars
# Enhancement: remove vars that were displayed ealier
dsp_numeric_feats_dstrb(feats_vctr=setdiff(names(glbObsAll),
c(myfind_chr_cols_df(glbObsAll), glb_rsp_var_raw, glb_rsp_var,
glbFeatsExclude)))
# Convert factors to dummy variables
# Build splines require(splines); bsBasis <- bs(training$age, df=3)
#pairs(subset(glbObsTrn, select=-c(col_symbol)))
# Check for glbObsNew & glbObsTrn features range mismatches
# Other diagnostics:
# print(subset(glbObsTrn, <col1_name> == max(glbObsTrn$<col1_name>, na.rm=TRUE) &
# <col2_name> <= mean(glbObsTrn$<col1_name>, na.rm=TRUE)))
# print(glbObsTrn[which.max(glbObsTrn$<col_name>),])
# print(<col_name>_freq_glbObsTrn <- mycreate_tbl_df(glbObsTrn, "<col_name>"))
# print(which.min(table(glbObsTrn$<col_name>)))
# print(which.max(table(glbObsTrn$<col_name>)))
# print(which.max(table(glbObsTrn$<col1_name>, glbObsTrn$<col2_name>)[, 2]))
# print(table(glbObsTrn$<col1_name>, glbObsTrn$<col2_name>))
# print(table(is.na(glbObsTrn$<col1_name>), glbObsTrn$<col2_name>))
# print(table(sign(glbObsTrn$<col1_name>), glbObsTrn$<col2_name>))
# print(mycreate_xtab_df(glbObsTrn, <col1_name>))
# print(mycreate_xtab_df(glbObsTrn, c(<col1_name>, <col2_name>)))
# print(<col1_name>_<col2_name>_xtab_glbObsTrn <-
# mycreate_xtab_df(glbObsTrn, c("<col1_name>", "<col2_name>")))
# <col1_name>_<col2_name>_xtab_glbObsTrn[is.na(<col1_name>_<col2_name>_xtab_glbObsTrn)] <- 0
# print(<col1_name>_<col2_name>_xtab_glbObsTrn <-
# mutate(<col1_name>_<col2_name>_xtab_glbObsTrn,
# <col3_name>=(<col1_name> * 1.0) / (<col1_name> + <col2_name>)))
# print(mycreate_sqlxtab_df(glbObsAll, c("<col1_name>", "<col2_name>")))
# print(<col2_name>_min_entity_arr <-
# sort(tapply(glbObsTrn$<col1_name>, glbObsTrn$<col2_name>, min, na.rm=TRUE)))
# print(<col1_name>_na_by_<col2_name>_arr <-
# sort(tapply(glbObsTrn$<col1_name>.NA, glbObsTrn$<col2_name>, mean, na.rm=TRUE)))
# Other plots:
# print(myplot_box(df=glbObsTrn, ycol_names="<col1_name>"))
# print(myplot_box(df=glbObsTrn, ycol_names="<col1_name>", xcol_name="<col2_name>"))
# print(myplot_line(subset(glbObsTrn, Symbol %in% c("CocaCola", "ProcterGamble")),
# "Date.POSIX", "StockPrice", facet_row_colnames="Symbol") +
# geom_vline(xintercept=as.numeric(as.POSIXlt("2003-03-01"))) +
# geom_vline(xintercept=as.numeric(as.POSIXlt("1983-01-01")))
# )
# print(myplot_line(subset(glbObsTrn, Date.POSIX > as.POSIXct("2004-01-01")),
# "Date.POSIX", "StockPrice") +
# geom_line(aes(color=Symbol)) +
# coord_cartesian(xlim=c(as.POSIXct("1990-01-01"),
# as.POSIXct("2000-01-01"))) +
# coord_cartesian(ylim=c(0, 250)) +
# geom_vline(xintercept=as.numeric(as.POSIXlt("1997-09-01"))) +
# geom_vline(xintercept=as.numeric(as.POSIXlt("1997-11-01")))
# )
# print(myplot_scatter(glbObsAll, "<col1_name>", "<col2_name>", smooth=TRUE))
# print(myplot_scatter(glbObsAll, "<col1_name>", "<col2_name>", colorcol_name="<Pred.fctr>") +
# geom_point(data=subset(glbObsAll, <condition>),
# mapping=aes(x=<x_var>, y=<y_var>), color="red", shape=4, size=5) +
# geom_vline(xintercept=84))
glb_chunks_df <- myadd_chunk(glb_chunks_df, "scrub.data", major.inc=FALSE)## label step_major step_minor label_minor bgn end elapsed
## 2 inspect.data 2 0 0 18.458 21.34 2.882
## 3 scrub.data 2 1 1 21.340 NA NAStep 2.1: scrub data
mycheck_problem_data(glbObsAll, featsExclude = glbFeatsExclude,
fctrMaxUniqVals = glbFctrMaxUniqVals)## [1] "numeric data missing in glbObsAll: "
## Popular Popular.fctr
## 1870 1870
## [1] "numeric data w/ 0s in glbObsAll: "
## WordCount Popular
## 109 5439
## [1] "numeric data w/ Infs in glbObsAll: "
## named integer(0)
## [1] "numeric data w/ NaNs in glbObsAll: "
## named integer(0)
## [1] "string data missing in glbObsAll: "
## NewsDesk SectionName SubsectionName Headline Snippet
## 2408 2899 6176 0 13
## Abstract PubDate
## 17 0findOffendingCharacter <- function(x, maxStringLength=256){
print(x)
for (c in 1:maxStringLength){
offendingChar <- substr(x,c,c)
#print(offendingChar) #uncomment if you want the indiv characters printed
#the next character is the offending multibyte Character
}
}
# string_vector <- c("test", "Se\x96ora", "works fine")
# lapply(string_vector, findOffendingCharacter)
# lapply(glbObsAll$description[29], findOffendingCharacter)
dsp_hdlxtab <- function(str)
print(mycreate_sqlxtab_df(glbObsAll[sel_obs(Headline.contains=str), ],
c("Headline.pfx", "Headline", glb_rsp_var)))
#dsp_hdlxtab("(1914)|(1939)")
dsp_catxtab <- function(str)
print(mycreate_sqlxtab_df(glbObsAll[sel_obs(Headline.contains=str), ],
c("Headline.pfx", "NewsDesk", "SectionName", "SubsectionName", glb_rsp_var)))
# dsp_catxtab("1914)|(1939)")
# dsp_catxtab("19(14|39|64):")
# dsp_catxtab("19..:")
# Merge some categories
# glbObsAll$myCategory <-
# plyr::revalue(glbObsAll$myCategory, c(
# "#Business Day#Dealbook" = "Business#Business Day#Dealbook",
# "#Business Day#Small Business" = "Business#Business Day#Small Business",
# "dummy" = "dummy"
# ))
# ctgry_xtab_df <- orderBy(reformulate(c("-", ".n")),
# mycreate_sqlxtab_df(glbObsAll,
# c("myCategory", "NewsDesk", "SectionName", "SubsectionName", glb_rsp_var)))
# myprint_df(ctgry_xtab_df)
# write.table(ctgry_xtab_df, paste0(glb_out_pfx, "ctgry_xtab.csv"),
# row.names=FALSE)
# ctgry_cast_df <- orderBy(~ -Y -NA, dcast(ctgry_xtab_df,
# myCategory + NewsDesk + SectionName + SubsectionName ~
# Popular.fctr, sum, value.var=".n"))
# myprint_df(ctgry_cast_df)
# write.table(ctgry_cast_df, paste0(glb_out_pfx, "ctgry_cast.csv"),
# row.names=FALSE)
# print(ctgry_sum_tbl <- table(glbObsAll$myCategory, glbObsAll[, glb_rsp_var],
# useNA="ifany"))
dsp_chisq.test <- function(...) {
sel_df <- glbObsAll[sel_obs(...) &
!is.na(glbObsAll$Popular), ]
sel_df$.marker <- 1
ref_df <- glbObsAll[!is.na(glbObsAll$Popular), ]
mrg_df <- merge(ref_df[, c(glb_id_var, "Popular")],
sel_df[, c(glb_id_var, ".marker")], all.x=TRUE)
mrg_df[is.na(mrg_df)] <- 0
print(mrg_tbl <- table(mrg_df$.marker, mrg_df$Popular))
print("Rows:Selected; Cols:Popular")
#print(mrg_tbl)
print(chisq.test(mrg_tbl))
}
# dsp_chisq.test(Headline.contains="[Ee]bola")
# dsp_chisq.test(Snippet.contains="[Ee]bola")
# dsp_chisq.test(Abstract.contains="[Ee]bola")
# print(mycreate_sqlxtab_df(glbObsAll[sel_obs(Headline.contains="[Ee]bola"), ],
# c(glb_rsp_var, "NewsDesk", "SectionName", "SubsectionName")))
# print(table(glbObsAll$NewsDesk, glbObsAll$SectionName))
# print(table(glbObsAll$SectionName, glbObsAll$SubsectionName))
# print(table(glbObsAll$NewsDesk, glbObsAll$SectionName, glbObsAll$SubsectionName))
# glbObsAll$myCategory.fctr <- as.factor(glbObsAll$myCategory)Step 2.1: scrub data
glb_chunks_df <- myadd_chunk(glb_chunks_df, "transform.data", major.inc=FALSE)## label step_major step_minor label_minor bgn end elapsed
## 3 scrub.data 2 1 1 21.340 22.44 1.1
## 4 transform.data 2 2 2 22.441 NA NA### Mapping dictionary
#sav_allobs_df <- glbObsAll; glbObsAll <- sav_allobs_df
if (!is.null(glb_map_vars)) {
for (feat in glb_map_vars) {
map_df <- myimport_data(url=glb_map_urls[[feat]],
comment="map_df",
print_diagn=TRUE)
glbObsAll <- mymap_codes(glbObsAll, feat, names(map_df)[2],
map_df, map_join_col_name=names(map_df)[1],
map_tgt_col_name=names(map_df)[2])
}
glbFeatsExclude <- union(glbFeatsExclude, glb_map_vars)
}
### Forced Assignments
#stop(here"); sav_allobs_df <- glbObsAll; glbObsAll <- sav_allobs_df
for (feat in glb_assign_vars) {
new_feat <- paste0(feat, ".my")
print(sprintf("Forced Assignments for: %s -> %s...", feat, new_feat))
glbObsAll[, new_feat] <- glbObsAll[, feat]
pairs <- glb_assign_pairs_lst[[feat]]
for (pair_ix in 1:length(pairs$from)) {
if (is.na(pairs$from[pair_ix]))
nobs <- nrow(filter(glbObsAll,
is.na(eval(parse(text=feat),
envir=glbObsAll)))) else
nobs <- sum(glbObsAll[, feat] == pairs$from[pair_ix])
#nobs <- nrow(filter(glbObsAll, is.na(Married.fctr))) ; print(nobs)
if ((is.na(pairs$from[pair_ix])) && (is.na(pairs$to[pair_ix])))
stop("what are you trying to do ???")
if (is.na(pairs$from[pair_ix]))
glbObsAll[is.na(glbObsAll[, feat]), new_feat] <-
pairs$to[pair_ix] else
glbObsAll[glbObsAll[, feat] == pairs$from[pair_ix], new_feat] <-
pairs$to[pair_ix]
print(sprintf(" %s -> %s for %s obs",
pairs$from[pair_ix], pairs$to[pair_ix], format(nobs, big.mark=",")))
}
glbFeatsExclude <- union(glbFeatsExclude, glb_assign_vars)
}
### Derivations using mapping functions
#stop(here"); sav_allobs_df <- glbObsAll; glbObsAll <- sav_allobs_df
for (new_feat in glb_derive_vars) {
print(sprintf("Creating new feature: %s...", new_feat))
args_lst <- NULL
for (arg in glbFeatsDerive[[new_feat]]$args)
args_lst[[arg]] <- glbObsAll[, arg]
glbObsAll[, new_feat] <- do.call(glbFeatsDerive[[new_feat]]$mapfn, args_lst)
}## [1] "Creating new feature: NDSSName.my..."
## [1] "Creating new feature: WordCount.log1p..."
## [1] "Creating new feature: WordCount.root2..."
## [1] "Creating new feature: WordCount.nexp..."#stop(here")
#hex_vctr <- c("\n", "\211", "\235", "\317", "\333")
hex_regex <- paste0(c("\n", "\211", "\235", "\317", "\333"), collapse="|")
for (obs_id in c(10029, 10948, 10136, 10178, 11514, 11904, 12157, 12210, 12659)) {
# tmp_str <- unlist(strsplit(glbObsAll[row_pos, "descr.my"], ""))
# glbObsAll[row_pos, "descr.my"] <- paste0(tmp_str[!tmp_str %in% hex_vctr],
# collapse="")
row_pos <- which(glbObsAll$UniqueID == obs_id)
# glbObsAll[row_pos, "descr.my"] <-
# gsub(hex_regex, " ", glbObsAll[row_pos, "descr.my"])
}Step 2.2: transform data
#```{r extract_features, cache=FALSE, eval=!is.null(glbFeatsText)}
glb_chunks_df <- myadd_chunk(glb_chunks_df, "extract.features", major.inc=TRUE)## label step_major step_minor label_minor bgn end elapsed
## 4 transform.data 2 2 2 22.441 22.531 0.091
## 5 extract.features 3 0 0 22.532 NA NAextract.features_chunk_df <- myadd_chunk(NULL, "extract.features_bgn")## label step_major step_minor label_minor bgn end
## 1 extract.features_bgn 1 0 0 22.539 NA
## elapsed
## 1 NA# Create new features that help prediction
# <col_name>.lag.2 <- lag(zoo(glbObsTrn$<col_name>), -2, na.pad=TRUE)
# glbObsTrn[, "<col_name>.lag.2"] <- coredata(<col_name>.lag.2)
# <col_name>.lag.2 <- lag(zoo(glbObsNew$<col_name>), -2, na.pad=TRUE)
# glbObsNew[, "<col_name>.lag.2"] <- coredata(<col_name>.lag.2)
#
# glbObsNew[1, "<col_name>.lag.2"] <- glbObsTrn[nrow(glbObsTrn) - 1,
# "<col_name>"]
# glbObsNew[2, "<col_name>.lag.2"] <- glbObsTrn[nrow(glbObsTrn),
# "<col_name>"]
# glbObsAll <- mutate(glbObsAll,
# A.P.http=ifelse(grepl("http",Added,fixed=TRUE), 1, 0)
# )
#
# glbObsTrn <- mutate(glbObsTrn,
# )
#
# glbObsNew <- mutate(glbObsNew,
# )
# Convert dates to numbers
# typically, dates come in as chars;
# so this must be done before converting chars to factors
#stop(here"); sav_allobs_df <- glbObsAll #; glbObsAll <- sav_allobs_df
if (!is.null(glb_date_vars)) {
glbObsAll <- cbind(glbObsAll,
myextract_dates_df(df=glbObsAll, vars=glb_date_vars,
id_vars=glb_id_var, rsp_var=glb_rsp_var))
for (sfx in c("", ".POSIX"))
glbFeatsExclude <-
union(glbFeatsExclude,
paste(glb_date_vars, sfx, sep=""))
for (feat in glb_date_vars) {
glbObsAll <- orderBy(reformulate(paste0(feat, ".POSIX")), glbObsAll)
# print(myplot_scatter(glbObsAll, xcol_name=paste0(feat, ".POSIX"),
# ycol_name=glb_rsp_var, colorcol_name=glb_rsp_var))
print(myplot_scatter(glbObsAll[glbObsAll[, paste0(feat, ".POSIX")] >=
strptime("2012-12-01", "%Y-%m-%d"), ],
xcol_name=paste0(feat, ".POSIX"),
ycol_name=glb_rsp_var, colorcol_name=paste0(feat, ".wkend")))
# Create features that measure the gap between previous timestamp in the data
require(zoo)
z <- zoo(as.numeric(as.POSIXlt(glbObsAll[, paste0(feat, ".POSIX")])))
glbObsAll[, paste0(feat, ".zoo")] <- z
print(head(glbObsAll[, c(glb_id_var, feat, paste0(feat, ".zoo"))]))
print(myplot_scatter(glbObsAll[glbObsAll[, paste0(feat, ".POSIX")] >
strptime("2012-10-01", "%Y-%m-%d"), ],
xcol_name=paste0(feat, ".zoo"), ycol_name=glb_rsp_var,
colorcol_name=glb_rsp_var))
b <- zoo(, seq(nrow(glbObsAll)))
last1 <- as.numeric(merge(z-lag(z, -1), b, all=TRUE)); last1[is.na(last1)] <- 0
glbObsAll[, paste0(feat, ".last1.log")] <- log(1 + last1)
print(gp <- myplot_box(df=glbObsAll[glbObsAll[,
paste0(feat, ".last1.log")] > 0, ],
ycol_names=paste0(feat, ".last1.log"),
xcol_name=glb_rsp_var))
last2 <- as.numeric(merge(z-lag(z, -2), b, all=TRUE)); last2[is.na(last2)] <- 0
glbObsAll[, paste0(feat, ".last2.log")] <- log(1 + last2)
print(gp <- myplot_box(df=glbObsAll[glbObsAll[,
paste0(feat, ".last2.log")] > 0, ],
ycol_names=paste0(feat, ".last2.log"),
xcol_name=glb_rsp_var))
last10 <- as.numeric(merge(z-lag(z, -10), b, all=TRUE)); last10[is.na(last10)] <- 0
glbObsAll[, paste0(feat, ".last10.log")] <- log(1 + last10)
print(gp <- myplot_box(df=glbObsAll[glbObsAll[,
paste0(feat, ".last10.log")] > 0, ],
ycol_names=paste0(feat, ".last10.log"),
xcol_name=glb_rsp_var))
last100 <- as.numeric(merge(z-lag(z, -100), b, all=TRUE)); last100[is.na(last100)] <- 0
glbObsAll[, paste0(feat, ".last100.log")] <- log(1 + last100)
print(gp <- myplot_box(df=glbObsAll[glbObsAll[,
paste0(feat, ".last100.log")] > 0, ],
ycol_names=paste0(feat, ".last100.log"),
xcol_name=glb_rsp_var))
glbObsAll <- orderBy(reformulate(glb_id_var), glbObsAll)
glbFeatsExclude <- union(glbFeatsExclude,
c(paste0(feat, ".zoo")))
# all2$last3 = as.numeric(merge(z-lag(z, -3), b, all = TRUE))
# all2$last5 = as.numeric(merge(z-lag(z, -5), b, all = TRUE))
# all2$last10 = as.numeric(merge(z-lag(z, -10), b, all = TRUE))
# all2$last20 = as.numeric(merge(z-lag(z, -20), b, all = TRUE))
# all2$last50 = as.numeric(merge(z-lag(z, -50), b, all = TRUE))
#
#
# # order table
# all2 = all2[order(all2$id),]
#
# ## fill in NAs
# # count averages
# na.avg = all2 %>% group_by(weekend, hour) %>% dplyr::summarise(
# last1=mean(last1, na.rm=TRUE),
# last3=mean(last3, na.rm=TRUE),
# last5=mean(last5, na.rm=TRUE),
# last10=mean(last10, na.rm=TRUE),
# last20=mean(last20, na.rm=TRUE),
# last50=mean(last50, na.rm=TRUE)
# )
#
# # fill in averages
# na.merge = merge(all2, na.avg, by=c("weekend","hour"))
# na.merge = na.merge[order(na.merge$id),]
# for(i in c("last1", "last3", "last5", "last10", "last20", "last50")) {
# y = paste0(i, ".y")
# idx = is.na(all2[[i]])
# all2[idx,][[i]] <- na.merge[idx,][[y]]
# }
# rm(na.avg, na.merge, b, i, idx, n, pd, sec, sh, y, z)
}
}
#rm(last1, last10, last100)
# Create factors of string variables
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "factorize.str.vars"), major.inc=TRUE)## label step_major step_minor label_minor
## 1 extract.features_bgn 1 0 0
## 2 extract.features_factorize.str.vars 2 0 0
## bgn end elapsed
## 1 22.539 22.552 0.013
## 2 22.552 NA NA#stop(here"); sav_allobs_df <- glbObsAll; #glbObsAll <- sav_allobs_df
print(str_vars <- myfind_chr_cols_df(glbObsAll))## NewsDesk SectionName SubsectionName Headline
## "NewsDesk" "SectionName" "SubsectionName" "Headline"
## Snippet Abstract PubDate .src
## "Snippet" "Abstract" "PubDate" ".src"
## NDSSName.my
## "NDSSName.my"if (length(str_vars <- setdiff(str_vars,
c(glbFeatsExclude, glbFeatsText))) > 0) {
for (var in str_vars) {
warning("Creating factors of string variable: ", var,
": # of unique values: ", length(unique(glbObsAll[, var])))
glbObsAll[, paste0(var, ".fctr")] <-
relevel(factor(glbObsAll[, var]),
names(which.max(table(glbObsAll[, var], useNA = "ifany"))))
}
glbFeatsExclude <- union(glbFeatsExclude, str_vars)
}## Warning: Creating factors of string variable: NDSSName.my: # of unique
## values: 21if (!is.null(glbFeatsText)) {
require(foreach)
require(gsubfn)
require(stringr)
require(tm)
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "process.text"), major.inc=TRUE)
chk_pattern_freq <- function(rex_str, ignore.case=TRUE) {
match_mtrx <- str_extract_all(txt_vctr, regex(rex_str, ignore_case=ignore.case),
simplify=TRUE)
match_df <- as.data.frame(match_mtrx[match_mtrx != ""])
names(match_df) <- "pattern"
return(mycreate_sqlxtab_df(match_df, "pattern"))
}
# match_lst <- gregexpr("\\bok(?!ay)", txt_vctr[746], ignore.case = FALSE, perl=TRUE); print(match_lst)
dsp_pattern <- function(rex_str, ignore.case=TRUE, print.all=TRUE) {
match_lst <- gregexpr(rex_str, txt_vctr, ignore.case = ignore.case, perl=TRUE)
match_lst <- regmatches(txt_vctr, match_lst)
match_df <- data.frame(matches=sapply(match_lst,
function (elems) paste(elems, collapse="#")))
match_df <- subset(match_df, matches != "")
if (print.all)
print(match_df)
return(match_df)
}
dsp_matches <- function(rex_str, ix) {
print(match_pos <- gregexpr(rex_str, txt_vctr[ix], perl=TRUE))
print(str_sub(txt_vctr[ix], (match_pos[[1]] / 100) * 99 + 0,
(match_pos[[1]] / 100) * 100 + 100))
}
myapply_gsub <- function(...) {
if ((length_lst <- length(names(gsub_map_lst))) == 0)
return(txt_vctr)
for (ptn_ix in 1:length_lst) {
if ((ptn_ix %% 10) == 0)
print(sprintf("running gsub for %02d (of %02d): #%s#...", ptn_ix,
length(names(gsub_map_lst)), names(gsub_map_lst)[ptn_ix]))
txt_vctr <- gsub(names(gsub_map_lst)[ptn_ix], gsub_map_lst[[ptn_ix]],
txt_vctr, ...)
}
return(txt_vctr)
}
myapply_txtmap <- function(txt_vctr, ...) {
nrows <- nrow(glb_txt_map_df)
for (ptn_ix in 1:nrows) {
if ((ptn_ix %% 10) == 0)
print(sprintf("running gsub for %02d (of %02d): #%s#...", ptn_ix,
nrows, glb_txt_map_df[ptn_ix, "rex_str"]))
txt_vctr <- gsub(glb_txt_map_df[ptn_ix, "rex_str"],
glb_txt_map_df[ptn_ix, "rpl_str"],
txt_vctr, ...)
}
return(txt_vctr)
#print(txt_vctr <- glbObsAll[glbObsAll$UniqueID == 11329, "descr.my"])
#strsplit(txt_vctr, "")[[1]][1]
#ptn_ix <- 2; glb_txt_map_df[ptn_ix, ]
#gsub(glb_txt_map_df[ptn_ix, "rex_str"], glb_txt_map_df[ptn_ix, "rpl_str"], txt_vctr)
#print(match_lst <- gregexpr(glb_txt_map_df[ptn_ix, "rex_str"], txt_vctr))
#strsplit(glb_txt_map_df[ptn_ix, "rex_str"], "")[[1]]
}
chk.equal <- function(bgn, end) {
print(all.equal(sav_txt_lst[["Headline"]][bgn:end],
glb_txt_chr_lst[["Headline"]][bgn:end]))
}
dsp.equal <- function(bgn, end) {
print(sav_txt_lst[["Headline"]][bgn:end])
print(glb_txt_chr_lst[["Headline"]][bgn:end])
}
#sav_txt_lst <- glb_txt_chr_lst; all.equal(sav_txt_lst, glb_txt_chr_lst)
#all.equal(sav_txt_lst[["Headline"]][1:4200], glb_txt_chr_lst[["Headline"]][1:4200])
#chk.equal( 1, 100)
#dsp.equal(86, 90)
txt_map_filename <- paste0(glb_txt_munge_filenames_pfx, "map.csv")
if (!file.exists(txt_map_filename))
stop(txt_map_filename, " not found!")
glb_txt_map_df <- read.csv(txt_map_filename, comment.char="#", strip.white=TRUE)
glb_txt_chr_lst <- list();
print(sprintf("Building glb_txt_chr_lst..."))
glb_txt_chr_lst <- foreach(txt_var = glbFeatsText) %dopar% {
# for (txt_var in glbFeatsText) {
txt_vctr <- glbObsAll[, txt_var]
names(txt_vctr) <- glbObsAll[, glb_id_var]
# myapply_txtmap shd be created as a tm_map::content_transformer ?
#print(glb_txt_map_df)
#txt_var=glbFeatsText[3]; txt_vctr <- glb_txt_chr_lst[[txt_var]]
#print(rex_str <- glb_txt_map_df[3, "rex_str"])
#print(rex_str <- glb_txt_map_df[glb_txt_map_df$rex_str == "\\bWall St\\.", "rex_str"])
#print(rex_str <- glb_txt_map_df[grepl("du Pont", glb_txt_map_df$rex_str), "rex_str"])
#print(rex_str <- glb_txt_map_df[glb_txt_map_df$rpl_str == "versus", "rex_str"])
#print(tmp_vctr <- grep(rex_str, txt_vctr, value=TRUE, ignore.case=FALSE))
#ret_lst <- regexec(rex_str, txt_vctr, ignore.case=FALSE); ret_lst <- regmatches(txt_vctr, ret_lst); ret_vctr <- sapply(1:length(ret_lst), function(pos_ix) ifelse(length(ret_lst[[pos_ix]]) > 0, ret_lst[[pos_ix]], "")); print(ret_vctr <- ret_vctr[ret_vctr != ""])
#gsub(rex_str, glb_txt_map_df[glb_txt_map_df$rex_str == rex_str, "rpl_str"], tmp_vctr, ignore.case=FALSE)
#grep("Hong Hong", txt_vctr, value=TRUE)
txt_vctr <- myapply_txtmap(txt_vctr, ignore.case=FALSE)
}
names(glb_txt_chr_lst) <- glbFeatsText
for (txt_var in glbFeatsText) {
print(sprintf("Remaining OK in %s:", txt_var))
txt_vctr <- glb_txt_chr_lst[[txt_var]]
print(chk_pattern_freq(rex_str <- "(?<!(BO|HO|LO))OK(?!(E\\!|ED|IE|IN|S ))",
ignore.case=FALSE))
match_df <- dsp_pattern(rex_str, ignore.case=FALSE, print.all=FALSE)
for (row in row.names(match_df))
dsp_matches(rex_str, ix=as.numeric(row))
print(chk_pattern_freq(rex_str <- "Ok(?!(a\\.|ay|in|ra|um))", ignore.case=FALSE))
match_df <- dsp_pattern(rex_str, ignore.case=FALSE, print.all=FALSE)
for (row in row.names(match_df))
dsp_matches(rex_str, ix=as.numeric(row))
print(chk_pattern_freq(rex_str <- "(?<!( b| B| c| C| g| G| j| M| p| P| w| W| r| Z|\\(b|ar|bo|Bo|co|Co|Ew|gk|go|ho|ig|jo|kb|ke|Ke|ki|lo|Lo|mo|mt|no|No|po|ra|ro|sm|Sm|Sp|to|To))ok(?!(ay|bo|e |e\\)|e,|e\\.|eb|ed|el|en|er|es|ey|i |ie|in|it|ka|ke|ki|ly|on|oy|ra|st|u |uc|uy|yl|yo))",
ignore.case=FALSE))
match_df <- dsp_pattern(rex_str, ignore.case=FALSE, print.all=FALSE)
for (row in row.names(match_df))
dsp_matches(rex_str, ix=as.numeric(row))
}
# txt_vctr <- glb_txt_chr_lst[[glbFeatsText[1]]]
# print(chk_pattern_freq(rex_str <- "(?<!( b| c| C| p|\\(b|bo|co|lo|Lo|Sp|to|To))ok(?!(ay|e |e\\)|e,|e\\.|ed|el|en|es|ey|ie|in|on|ra))", ignore.case=FALSE))
# print(chk_pattern_freq(rex_str <- "ok(?!(ay|el|on|ra))", ignore.case=FALSE))
# dsp_pattern(rex_str, ignore.case=FALSE, print.all=FALSE)
# dsp_matches(rex_str, ix=8)
# substr(txt_vctr[86], 5613, 5620)
# substr(glbObsAll[301, "review"], 550, 650)
#stop(here"); sav_txt_lst <- glb_txt_chr_lst
for (txt_var in glbFeatsText) {
print(sprintf("Remaining Acronyms in %s:", txt_var))
txt_vctr <- glb_txt_chr_lst[[txt_var]]
print(chk_pattern_freq(rex_str <- "([[:upper:]]\\.( *)){2,}", ignore.case=FALSE))
# Check for names
print(subset(chk_pattern_freq(rex_str <- "(([[:upper:]]+)\\.( *)){1}",
ignore.case=FALSE),
.n > 1))
# dsp_pattern(rex_str="(OK\\.( *)){1}", ignore.case=FALSE)
# dsp_matches(rex_str="(OK\\.( *)){1}", ix=557)
#dsp_matches(rex_str="\\bR\\.I\\.P(\\.*)(\\B)", ix=461)
#dsp_matches(rex_str="\\bR\\.I\\.P(\\.*)", ix=461)
#print(str_sub(txt_vctr[676], 10100, 10200))
#print(str_sub(txt_vctr[74], 1, -1))
}
for (txt_var in glbFeatsText) {
re_str <- "\\b(Fort|Ft\\.|Hong|Las|Los|New|Puerto|Saint|San|St\\.)( |-)(\\w)+"
print(sprintf("Remaining #%s# terms in %s: ", re_str, txt_var))
txt_vctr <- glb_txt_chr_lst[[txt_var]]
print(orderBy(~ -.n +pattern, subset(chk_pattern_freq(re_str, ignore.case=FALSE),
grepl("( |-)[[:upper:]]", pattern))))
print(" consider cleaning if relevant to problem domain; geography name; .n > 1")
#grep("New G", txt_vctr, value=TRUE, ignore.case=FALSE)
#grep("St\\. Wins", txt_vctr, value=TRUE, ignore.case=FALSE)
}
#stop(here"); sav_txt_lst <- glb_txt_chr_lst
for (txt_var in glbFeatsText) {
re_str <- "\\b(N|S|E|W|C)( |\\.)(\\w)+"
print(sprintf("Remaining #%s# terms in %s: ", re_str, txt_var))
txt_vctr <- glb_txt_chr_lst[[txt_var]]
print(orderBy(~ -.n +pattern, subset(chk_pattern_freq(re_str, ignore.case=FALSE),
grepl(".", pattern))))
#grep("N Weaver", txt_vctr, value=TRUE, ignore.case=FALSE)
}
for (txt_var in glbFeatsText) {
re_str <- "\\b(North|South|East|West|Central)( |\\.)(\\w)+"
print(sprintf("Remaining #%s# terms in %s: ", re_str, txt_var))
txt_vctr <- glb_txt_chr_lst[[txt_var]]
if (nrow(filtered_df <- subset(chk_pattern_freq(re_str, ignore.case=FALSE),
grepl(".", pattern))) > 0)
print(orderBy(~ -.n +pattern, filtered_df))
#grep("Central (African|Bankers|Cast|Italy|Role|Spring)", txt_vctr, value=TRUE, ignore.case=FALSE)
#grep("East (Africa|Berlin|London|Poland|Rivals|Spring)", txt_vctr, value=TRUE, ignore.case=FALSE)
#grep("North (American|Korean|West)", txt_vctr, value=TRUE, ignore.case=FALSE)
#grep("South (Pacific|Street)", txt_vctr, value=TRUE, ignore.case=FALSE)
#grep("St\\. Martins", txt_vctr, value=TRUE, ignore.case=FALSE)
}
find_cmpnd_wrds <- function(txt_vctr) {
# Enhancements:
# - arg should be txt_corpus instead of txt_vctr
txt_corpus <- Corpus(VectorSource(txt_vctr))
txt_corpus <- tm_map(txt_corpus, content_transformer(tolower), lazy = TRUE)
txt_corpus <- tm_map(txt_corpus, PlainTextDocument, lazy = TRUE)
txt_corpus <- tm_map(txt_corpus, removePunctuation,
preserve_intra_word_dashes = TRUE, lazy = FALSE)
# Defaulting to Tf since TfIdf with normalize = TRUE throws a warning for empty docs
terms_mtrx <- as.matrix(TermDocumentMatrix(txt_corpus,
control = list(weighting = weightTf)))
terms_df <- orderBy(~ -Tf, data.frame(term = dimnames(terms_mtrx)$Terms,
Tf = rowSums(terms_mtrx)))
cmpnd_df <- subset(terms_df, grepl("-", term))
if (nrow(cmpnd_df) == 0) {
print(" No compounded terms found")
return(FALSE)
}
txt_compound_filename <- paste0(glb_txt_munge_filenames_pfx, "compound.csv")
if (!file.exists(txt_compound_filename))
stop(txt_compound_filename, " not found!")
filter_df <- read.csv(txt_compound_filename, comment.char="#", strip.white=TRUE)
cmpnd_df$filter <- FALSE
for (row_ix in 1:nrow(filter_df))
cmpnd_df[!cmpnd_df$filter, "filter"] <-
grepl(filter_df[row_ix, "rex_str"],
cmpnd_df[!cmpnd_df$filter, "term"], ignore.case=TRUE)
cmpnd_df <- subset(cmpnd_df, !filter)
# Bug in tm_map(txt_corpus, removePunctuation, preserve_intra_word_dashes=TRUE) ???
# "net-a-porter" gets converted to "net-aporter"
#grep("net-a-porter", txt_vctr, ignore.case=TRUE, value=TRUE)
#grep("maser-laser", txt_vctr, ignore.case=TRUE, value=TRUE)
#txt_corpus[[which(grepl("net-a-porter", txt_vctr, ignore.case=TRUE))]]
#grep("\\b(across|longer)-(\\w)", cmpnd_df$term, ignore.case=TRUE, value=TRUE)
#grep("(\\w)-(affected|term)\\b", cmpnd_df$term, ignore.case=TRUE, value=TRUE)
print(sprintf("nrow(cmpnd_df): %d", nrow(cmpnd_df)))
myprint_df(cmpnd_df)
}
# This should be run after glb_txt_corpus_lst is created with tolower
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "process.text_reporting_compound_terms"), major.inc=FALSE)
# for (txt_var in glbFeatsText) {
# print(sprintf("Remaining compound terms in %s: ", txt_var))
# find_cmpnd_wrds(txt_vctr = glb_txt_chr_lst[[txt_var]])
# #grep("thirty-five", txt_vctr, ignore.case=TRUE, value=TRUE)
# #rex_str <- glb_txt_map_df[grepl("hirty", glb_txt_map_df$rex_str), "rex_str"]
# }
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "build.corpus"), major.inc=TRUE)
get_txt_terms <- function(terms_TDM) {
terms_mtrx <- as.matrix(as.TermDocumentMatrix(terms_TDM))
docms_mtrx <- as.matrix(as.DocumentTermMatrix(terms_TDM))
terms_df <- data.frame(term = dimnames(terms_mtrx)$Terms,
weight = rowSums(terms_mtrx),
freq = rowSums(terms_mtrx > 0))
terms_df$pos <- 1:nrow(terms_df)
terms_df$cor.y <-
cor(docms_mtrx[glbObsAll$.src == "Train",],
as.numeric(glbObsAll[glbObsAll$.src == "Train", glb_rsp_var]),
use = "pairwise.complete.obs")
terms_df$cor.y.abs <- abs(terms_df$cor.y)
# .rnorm.cor.y.abs <- abs(cor(glbObsAll[glbObsAll$.src == "Train", ".rnorm"],
# as.numeric(glbObsAll[glbObsAll$.src == "Train", glb_rsp_var]),
# use = "pairwise.complete.obs"))
terms_df$chisq.stat <- NA; terms_df$chisq.pval <- NA
for (ix in 1:nrow(terms_df)) {
#for (ix in 1:743) {
if (length(unique(docms_mtrx[glbObsAll$.src == "Train", ix])) > 1) {
chisq <- chisq.test(docms_mtrx[glbObsAll$.src == "Train", ix],
glbObsAll[glbObsAll$.src == "Train", glb_rsp_var])
terms_df[ix, "chisq.stat"] <- chisq$statistic
terms_df[ix, "chisq.pval"] <- chisq$p.value
}
}
nzv_df <- nzv(docms_mtrx[glbObsAll$.src == "Train",], freqCut = glb_nzv_freqCut,
uniqueCut = glb_nzv_uniqueCut, saveMetrics = TRUE)
terms_df$nzv.freqRatio <- nzv_df$freqRatio
# terms_df$nzv.freqRatio.cut.fctr <- cut(terms_df$nzv.freqRatio,
# breaks = sort(c(min(terms_df$nzv.freqRatio),
# glb_nzv_freqCut,
# max(terms_df$nzv.freqRatio))))
terms_df$nzv.percentUnique <- nzv_df$percentUnique
# terms_df$nzv.percentUnique.cut.fctr <- cut(terms_df$nzv.percentUnique,
# breaks = sort(c(min(terms_df$nzv.percentUnique) - .Machine$double.neg.eps,
# glb_nzv_uniqueCut,
# max(terms_df$nzv.percentUnique))))
# terms_df$nzv.quad.fctr <- as.factor(paste0("fRatio:", terms_df$nzv.freqRatio.cut.fctr,
# "\n%Unq:", terms_df$nzv.percentUnique.cut.fctr))
terms_df$nzv <- nzv_df$nzv
for (cls in unique(glbObsAll[, glb_txt_cor_var])) {
if (!is.na(cls))
terms_df[, paste0("weight.", as.character(cls))] <-
colSums(t(terms_mtrx) *
as.numeric(!is.na(glbObsAll[, glb_txt_cor_var]) &
(glbObsAll[, glb_txt_cor_var] == cls))) else
terms_df[, paste0("weight.", as.character(cls))] <-
colSums(t(terms_mtrx) *
as.numeric(is.na(glbObsAll[, glb_txt_cor_var])))
}
# Check all calls to get_terms_DTM_terms to change returned order assumption
return(terms_df <- orderBy(~ -weight, terms_df))
}
#plt_full_df <- get_terms_DTM_terms(terms_DTM=glb_full_terms_DTM_lst[[txt_var]])
get_corpus_terms <- function(txt_corpus) {
return(terms_df <- get_txt_terms(terms_TDM =
TermDocumentMatrix(txt_corpus, control = glb_txt_terms_control)))
}
myreplacePunctuation <- function(x) {
return(gsub("[[:punct:]]+", " ", gsub("('|-)", "", x)))
}
#stop(here"); glb_to_sav()
glb_txt_corpus_lst <- list()
print(sprintf("Building glb_txt_corpus_lst..."))
glb_txt_corpus_lst <- foreach(txt_var = glbFeatsText, .verbose = FALSE) %dopar% {
#glb_txt_corpus_lst <- foreach(txt_var = glbFeatsText, .verbose = TRUE) %do% {
#for (txt_var in glbFeatsText) {
txt_corpus <- Corpus(VectorSource(glb_txt_chr_lst[[txt_var]]))
txt_corpus <- tm_map(txt_corpus, PlainTextDocument, lazy = TRUE)
txt_corpus <- tm_map(txt_corpus, content_transformer(tolower), lazy = TRUE) #nuppr
# removePunctuation does not replace with whitespace. Use a custom transformer ???
#txt_corpus <- tm_map(txt_corpus, removePunctuation, lazy = TRUE) #npnct<chr_ix>
txt_corpus <- tm_map(txt_corpus, content_transformer(myreplacePunctuation)
, lazy = TRUE) #npnct<chr_ix>
# txt-corpus <- tm_map(txt_corpus, content_transformer(function(x, pattern) gsub(pattern, "", x))
if (!is.null(glb_txt_stop_words[[txt_var]]))
txt_corpus <- tm_map(txt_corpus, removeWords, glb_txt_stop_words[[txt_var]],
lazy = FALSE)#, lazy=TRUE) #nstopwrds
# foreach result is based on .Last.Eval
txt_corpus <- txt_corpus
# glb_txt_corpus_lst[[txt_var]] <- txt_corpus
}
names(glb_txt_corpus_lst) <- glbFeatsText
mycombineSynonyms <- content_transformer(function(x, syn=NULL) {
Reduce(function(a,b) {
gsub(paste0("\\b(", paste(b$syns, collapse = "|"),")\\b"), b$word, a)}, syn, x)
})
#stop(here"); glb_to_sav(); sav_txt_corpus <- glb_txt_corpus_lst[[txt_var]]; all.equal(sav_txt_corpus, glb_txt_corpus_lst[[txt_var]]); glb_txt_corpus_lst[[txt_var]] <- sav_txt_corpus
glb_post_stop_words_terms_df_lst <- list();
glb_post_stop_words_terms_mtrx_lst <- list();
glb_post_stem_words_terms_df_lst <- list();
glb_post_stem_words_terms_mtrx_lst <- list();
for (txt_var in glbFeatsText) {
print(sprintf(" Top_n post-stop term weights for %s:", txt_var))
# This impacts stemming probably due to lazy parameter
print(myprint_df(full_terms_df <-
get_corpus_terms(txt_corpus=glb_txt_corpus_lst[[txt_var]]),
glbFeatsTextTermsMax[[txt_var]]))
glb_post_stop_words_terms_df_lst[[txt_var]] <- full_terms_df
terms_stop_mtrx <- as.matrix(DocumentTermMatrix(glb_txt_corpus_lst[[txt_var]],
control=glb_txt_terms_control))
rownames(terms_stop_mtrx) <- rownames(glbObsAll) # print undreadable otherwise
glb_post_stop_words_terms_mtrx_lst[[txt_var]] <- terms_stop_mtrx
tmp_allobs_df <- glbObsAll[, c(glb_id_var, glb_rsp_var)]
tmp_allobs_df$terms.post.stop.n <- rowSums(terms_stop_mtrx > 0)
tmp_allobs_df$terms.post.stop.n.log <- log(1 + tmp_allobs_df$terms.post.stop.n)
tmp_allobs_df$weight.post.stop.sum <- rowSums(terms_stop_mtrx)
print(sprintf(" Top_n stem term weights for %s:", txt_var))
glb_txt_corpus_lst[[txt_var]] <- tm_map(glb_txt_corpus_lst[[txt_var]], stemDocument,
"english", lazy=FALSE)
if (!is.null(glb_txt_synonyms[[txt_var]])) {
syn_lst <- myrmNullObj(glb_txt_synonyms[[txt_var]])
glb_txt_corpus_lst[[txt_var]] <- tm_map(glb_txt_corpus_lst[[txt_var]],
mycombineSynonyms,
syn_lst, lazy=FALSE)
}
print(myprint_df(full_terms_df <- get_corpus_terms(glb_txt_corpus_lst[[txt_var]]),
glbFeatsTextTermsMax[[txt_var]]))
glb_post_stem_words_terms_df_lst[[txt_var]] <- full_terms_df
terms_stem_mtrx <- as.matrix(DocumentTermMatrix(glb_txt_corpus_lst[[txt_var]],
control=glb_txt_terms_control))
rownames(terms_stem_mtrx) <- rownames(glbObsAll) # print undreadable otherwise
glb_post_stem_words_terms_mtrx_lst[[txt_var]] <- terms_stem_mtrx
tmp_allobs_df$terms.post.stem.n <- rowSums(terms_stem_mtrx > 0)
tmp_allobs_df$terms.post.stem.n.log <- log(1 + tmp_allobs_df$terms.post.stem.n)
tmp_allobs_df$weight.post.stem.sum <- rowSums(terms_stem_mtrx)
tmp_allobs_df$terms.n.stem.stop.Ratio <-
1.0 * tmp_allobs_df$terms.post.stem.n / tmp_allobs_df$terms.post.stop.n
tmp_allobs_df[(is.nan(tmp_allobs_df$terms.n.stem.stop.Ratio) |
is.infinite(tmp_allobs_df$terms.n.stem.stop.Ratio)),
"terms.n.stem.stop.Ratio"] <- 1.0
if ((n.errors <- sum(tmp_allobs_df$terms.n.stem.stop.Ratio > 1)) > 0)
stop(n.errors, " obs in tmp_allobs_df have terms.n.stem.stop.Ratio > 1",
" happening due to terms filtered by glb_txt_terms_control$bounds$global[1] but stemmable to other terms")
#print(head(subset(tmp_allobs_df, terms.n.stem.stop.Ratio > 1)))
#glbObsAll[(row_ix <- which(glbObsAll$UniqueID == 10465)), ]
#terms_stop_mtrx[row_ix, terms_stop_mtrx[row_ix, ] > 0]
#setdiff(names(terms_stem_mtrx[row_ix, terms_stem_mtrx[row_ix, ] > 0]), names(terms_stop_mtrx[row_ix, terms_stop_mtrx[row_ix, ] > 0]))
#mydspObs(list(descr.my.contains="updat"))
tmp_allobs_df$weight.sum.stem.stop.Ratio <-
1.0 * tmp_allobs_df$weight.post.stem.sum / tmp_allobs_df$weight.post.stop.sum
tmp_allobs_df[is.nan(tmp_allobs_df$weight.sum.stem.stop.Ratio) |
is.infinite(tmp_allobs_df$weight.sum.stem.stop.Ratio),
"weight.sum.stem.stop.Ratio"] <- 1.0
tmp_trnobs_df <- tmp_allobs_df[!is.na(tmp_allobs_df[, glb_rsp_var]), ]
print(cor(as.matrix(tmp_trnobs_df[, -c(1, 2)]),
as.numeric(tmp_trnobs_df[, glb_rsp_var])))
txt_var_pfx <- toupper(substr(txt_var, 1, 1))
tmp_allobs_df <- tmp_allobs_df[, -c(1, 2)]
names(tmp_allobs_df) <- paste(paste0(txt_var_pfx, "."), names(tmp_allobs_df), sep = "")
glbObsAll <- cbind(glbObsAll, tmp_allobs_df)
glbFeatsExclude <- c(glbFeatsExclude,
paste(paste0(txt_var_pfx, ".terms.post."), c("stop.n", "stem.n"), sep = ""))
}
require(wordcloud)
for (txt_var in glbFeatsText) {
print(sprintf(" Wordcloud post-stem term weights for %s:", txt_var))
m <- glb_post_stem_words_terms_mtrx_lst[[txt_var]]
# calculate the frequency of words
v <- sort(colSums(m), decreasing = TRUE)
myNames <- names(v)
d <- data.frame(word = myNames, freq = v)
wordcloud(d$word, d$freq, min.freq = glb_txt_terms_control$bounds$global[1])
}
for (txt_var in glbFeatsText) {
.rnorm.cor.y.abs <- abs(cor(glbObsAll[glbObsAll$.src == "Train", ".rnorm"],
as.numeric(glbObsAll[glbObsAll$.src == "Train", glb_rsp_var]),
use = "pairwise.complete.obs"))
plt_df <- subset(glb_post_stem_words_terms_df_lst[[txt_var]], !is.na(cor.y))
plt_df$nzv.freqRatio.cut.fctr <- cut(plt_df$nzv.freqRatio,
breaks = sort(c(min(plt_df$nzv.freqRatio),
glb_nzv_freqCut,
max(plt_df$nzv.freqRatio))))
plt_df$nzv.percentUnique.cut.fctr <- cut(plt_df$nzv.percentUnique,
breaks = sort(c(min(plt_df$nzv.percentUnique) - .Machine$double.neg.eps,
glb_nzv_uniqueCut,
max(plt_df$nzv.percentUnique))))
plt_df$nzv.quad.fctr <- as.factor(paste0("fRatio:", plt_df$nzv.freqRatio.cut.fctr,
"\n%Unq:", plt_df$nzv.percentUnique.cut.fctr))
labelCnd <- !plt_df$nzv |
(!is.na(plt_df$chisq.pval) & (plt_df$chisq.pval < 0.05)) &
(!is.na(plt_df$cor.y.abs) & (plt_df$cor.y.abs > .rnorm.cor.y.abs))
plt_df$label <- NA; plt_df[labelCnd, "label"] <- plt_df[labelCnd, "term"]
print(ggplot(plt_df, aes(x = cor.y, y = chisq.stat)) +
geom_point(aes(color = nzv.quad.fctr, size = weight)) +
geom_text(aes(label = label), color = "gray50") +
# geom_vline(xintercept = 0) +
geom_vline(xintercept = c(-1, +1) * .rnorm.cor.y.abs, color = "gray",
linetype = "dashed") +
ggtitle(txt_var))
}
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "extract.DTM"), major.inc=TRUE)
glb_full_DTM_lst <- list(); glb_sprs_DTM_lst <- list();
for (txt_var in glbFeatsText) {
print(sprintf("Extracting term weights for %s...", txt_var))
txt_corpus <- glb_txt_corpus_lst[[txt_var]]
full_DTM <- DocumentTermMatrix(txt_corpus,
control=glb_txt_terms_control)
sprs_DTM <- removeSparseTerms(full_DTM,
glb_sprs_thresholds[txt_var])
glb_full_DTM_lst[[txt_var]] <- full_DTM
glb_sprs_DTM_lst[[txt_var]] <- sprs_DTM
}
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "report.DTM"), major.inc=TRUE)
require(reshape2)
for (txt_var in glbFeatsText) {
print(sprintf("Reporting term weights for %s...", txt_var))
full_DTM <- glb_full_DTM_lst[[txt_var]]
sprs_DTM <- glb_sprs_DTM_lst[[txt_var]]
print(" Full TermMatrix:"); print(full_DTM)
full_terms_df <- get_txt_terms(full_DTM)
# full_terms_df <- full_terms_df[, c(2, 1, 3, 4)]
# col_names <- names(full_terms_df)
# col_names[2:length(col_names)] <-
# paste(col_names[2:length(col_names)], ".full", sep="")
# names(full_terms_df) <- col_names
print(" Sparse TermMatrix:"); print(sprs_DTM)
sprs_terms_df <- get_txt_terms(sprs_DTM)
# sprs_terms_df <- sprs_terms_df[, c(2, 1, 3, 4)]
# col_names <- names(sprs_terms_df)
# col_names[2:length(col_names)] <-
# paste(col_names[2:length(col_names)], ".sprs", sep="")
# names(sprs_terms_df) <- col_names
#intersect(names(full_terms_df), names(sprs_terms_df))
terms_df <- merge(full_terms_df, sprs_terms_df, by = c("term", "weight", "freq",
grep("weight\\.", names(full_terms_df), value = TRUE)),
all.x = TRUE, suffixes = c(".full", ".sprs"))
terms_df$in.sprs <- !is.na(terms_df$pos.sprs)
plt_terms_df <- subset(terms_df,
weight >= min(terms_df$weight[!is.na(terms_df$pos.sprs)], na.rm=TRUE))
plt_terms_df$label <- ""
plt_terms_df[is.na(plt_terms_df$pos.sprs), "label"] <-
plt_terms_df[is.na(plt_terms_df$pos.sprs), "term"]
# glb_important_terms[[txt_var]] <- union(glb_important_terms[[txt_var]],
# plt_terms_df[is.na(plt_terms_df$TfIdf.sprs), "term"])
print(myplot_scatter(plt_terms_df, "freq", "weight",
colorcol_name="in.sprs") +
geom_text(aes(label=label), color="Black", size=3.5))
melt_terms_df <- orderBy(~ -value,
melt(terms_df, id.vars="term", measure.vars = c("weight", "freq")))
print(ggplot(melt_terms_df, aes(value, color=variable)) + stat_ecdf() +
geom_hline(yintercept=glb_sprs_thresholds[txt_var],
linetype = "dotted"))
melt_terms_df <- orderBy(~ -value,
melt(subset(terms_df, in.sprs), id.vars="term",
measure.vars=grep("weight.", names(terms_df), value=TRUE)))
print(myplot_hbar(melt_terms_df, "term", "value", colorcol_name="variable"))
melt_terms_df <- orderBy(~ -value,
melt(subset(terms_df, !in.sprs), id.vars="term",
measure.vars=grep("weight.", names(terms_df), value=TRUE)))
print(myplot_hbar(head(melt_terms_df, glbFeatsTextTermsMax[[txt_var]]), "term", "value",
colorcol_name="variable"))
}
# sav_full_DTM_lst <- glb_full_DTM_lst
# print(identical(sav_glb_txt_corpus_lst, glb_txt_corpus_lst))
# print(all.equal(length(sav_glb_txt_corpus_lst), length(glb_txt_corpus_lst)))
# print(all.equal(names(sav_glb_txt_corpus_lst), names(glb_txt_corpus_lst)))
# print(all.equal(sav_glb_txt_corpus_lst[["Headline"]], glb_txt_corpus_lst[["Headline"]]))
# print(identical(sav_full_DTM_lst, glb_full_DTM_lst))
rm(full_terms_mtrx)
# Create txt features
if ((length(glbFeatsText) > 1) &&
(length(unique(pfxs <- sapply(glbFeatsText,
function(txt) toupper(substr(txt, 1, 1))))) < length(glbFeatsText)))
stop("Prefixes for corpus freq terms not unique: ", pfxs)
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "bind.DTM"),
major.inc=TRUE)
#stop(here"); glb_to_sav(); all.equal(sav_allobs_df, glbObsAll); glbObsAll <- sav_allobs_df
require(tidyr)
for (txt_var in glbFeatsText) {
print(sprintf("Binding DTM for %s...", txt_var))
txt_var_pfx <- toupper(substr(txt_var, 1, 1))
txt_full_X_df <- as.data.frame(as.matrix(glb_full_DTM_lst[[txt_var]]))
terms_full_df <- get_txt_terms(glb_full_DTM_lst[[txt_var]])
# make.names adds a period to R keywords e.g. "in", "function"
colnames(txt_full_X_df) <- paste(txt_var_pfx, ".T.",
make.names(colnames(txt_full_X_df)), sep="")
rownames(txt_full_X_df) <- rownames(glbObsAll) # warning otherwise
# plt_full_df <- terms_full_df
# names(plt_full_df)[grepl("weight$", names(plt_full_df))] <- "weight.all"
# # gather(plt_full_df[1:5, ], domain, TfIdf, -matches("!(TfIdf)"))
# # gather(plt_full_df[1:5, grepl("TfIdf", names(plt_full_df))], domain, TfIdf)
# # gather(plt_full_df[1:5, ], domain, TfIdf,
# # -names(plt_full_df)[!grepl("TfIdf", names(plt_full_df))])
# plt_full_df <- gather(plt_full_df, domain, weight,
# -c(term, freq, pos, cor.y, cor.y.abs))
# plt_full_df$label <- NA
# top_val_terms <- orderBy(~-weight, terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]]
# plt_full_df[plt_full_df$term %in% top_val_terms, "label"] <-
# plt_full_df[plt_full_df$term %in% top_val_terms, "term"]
# top_cor_terms <- orderBy(~-cor.y.abs,
# terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]]
# plt_full_df[plt_full_df$term %in% top_cor_terms, "label"] <-
# plt_full_df[plt_full_df$term %in% top_cor_terms, "term"]
# #plt_full_df$type <- "none"
# plt_full_df[plt_full_df$term %in% top_val_terms, "type"] <- "top.weight"
# plt_full_df[plt_full_df$term %in% top_cor_terms, "type"] <- "top.cor"
# plt_full_df[plt_full_df$term %in% intersect(top_val_terms, top_cor_terms), "type"] <-
# "top.both"
# cor.y.rnorm <- cor(glbObsAll$.rnorm, as.numeric(glbObsAll[, glb_rsp_var]),
# use = "pairwise.complete.obs")
# print(ggplot(plt_full_df, aes(x=weight, y=cor.y)) + facet_wrap(~ domain) +
# geom_point(aes(size=freq), color="grey") +
# geom_jitter() +
# geom_text(aes(label=label, color=type), size=3.5) +
# #geom_hline(yintercept=cor.y.rnorm, color="red") +
# geom_hline(yintercept=c(cor.y.rnorm, -cor.y.rnorm), color="red"))
#stop(here"); glb_to_sav()
if (glbFeatsTextFilter == "sparse") {
txt_X_df <- as.data.frame(as.matrix(glb_sprs_DTM_lst[[txt_var]]))
select_terms <- make.names(colnames(txt_X_df))
# colnames(txt_X_df) <- paste(txt_var_pfx, ".T.",
# make.names(colnames(txt_X_df)), sep="")
# rownames(txt_X_df) <- rownames(glbObsAll) # warning otherwise
} else if (glbFeatsTextFilter == "top.val") {
select_terms <- orderBy(~-weight,
terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]]
# txt_X_df <- txt_full_X_df[, subset(terms_full_df, term %in% select_terms)$pos,
# FALSE]
} else if (glbFeatsTextFilter == "top.cor") {
select_terms <- orderBy(~-cor.y.abs,
terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]]
# txt_X_df <- txt_full_X_df[, subset(terms_full_df, term %in% select_terms)$pos,
# FALSE]
} else if (glbFeatsTextFilter == "top.chisq") {
select_terms <- orderBy(~-chisq.stat,
subset(terms_full_df, chisq.pval < 0.05)
)$term[1:glbFeatsTextTermsMax[[txt_var]]]
} else if (glbFeatsTextFilter == "union.top.val.cor") {
select_terms <- union(
orderBy(~-weight , terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]],
orderBy(~-cor.y.abs, terms_full_df)$term[1:glbFeatsTextTermsMax[[txt_var]]])
} else stop(
"glbFeatsTextFilter should be one of c('sparse', 'top.val', 'top.cor', 'union.top.val.cor', 'top.chisq') vs. '",
glbFeatsTextFilter, "'")
assoc_terms_lst <- findAssocs(glb_full_DTM_lst[[txt_var]], select_terms,
glbFeatsTextAssocCor[[txt_var]])
assoc_terms <- c(NULL)
for (term in names(assoc_terms_lst))
if (length(assoc_terms_lst[[term]]) > 0)
assoc_terms <- union(assoc_terms, names(assoc_terms_lst[[term]]))
#stop(here"); glb_to_sav()
txt_X_df <- txt_full_X_df[,
subset(terms_full_df, term %in% c(select_terms, assoc_terms))$pos,
FALSE]
glbObsAll <- cbind(glbObsAll, txt_X_df) # TfIdf is normalized
#glbObsAll <- cbind(glbObsAll, log_X_df) # if using non-normalized metrics
}
#identical(chk_entity_df, glbObsAll)
#chk_entity_df <- glbObsAll
extract.features_chunk_df <- myadd_chunk(extract.features_chunk_df,
paste0("extract.features_", "bind.DXM"),
major.inc=TRUE)
#stop(here"); sav_allobs_df <- glbObsAll; glbObsAll <- sav_allobs_df
glb_punct_vctr <- c("!", "\"", "#", "\\$", "%", "&", "'",
"\\(|\\)",# "\\(", "\\)",
"\\*", "\\+", ",", "-", "\\.", "/", ":", ";",
"<|>", # "<",
"=",
# ">",
"\\?", "@", "\\[", "\\\\", "\\]", "\\^", "_", "`",
"\\{", "\\|", "\\}", "~")
txt_X_df <- glbObsAll[, c(glb_id_var, ".rnorm"), FALSE]
txt_X_df <- foreach(txt_var=glbFeatsText, .combine=cbind) %dopar% {
#for (txt_var in glbFeatsText) {
print(sprintf("Binding DXM for %s...", txt_var))
txt_var_pfx <- toupper(substr(txt_var, 1, 1))
txt_full_DTM_mtrx <- as.matrix(glb_full_DTM_lst[[txt_var]])
rownames(txt_full_DTM_mtrx) <- rownames(glbObsAll) # print undreadable otherwise
#print(txt_full_DTM_mtrx[txt_full_DTM_mtrx[, "ebola"] != 0, "ebola"])
# Create <txt_var>.T.<term> for glb_important_terms
for (term in glb_important_terms[[txt_var]])
txt_X_df[, paste0(txt_var_pfx, ".T.", make.names(term))] <-
txt_full_DTM_mtrx[, term]
# Create <txt_var>.wrds.n.log & .wrds.unq.n.log
txt_X_df[, paste0(txt_var_pfx, ".wrds.n.log")] <-
log(1 + mycount_pattern_occ("\\w+", glb_txt_chr_lst[[txt_var]]))
txt_X_df[, paste0(txt_var_pfx, ".wrds.unq.n.log")] <-
log(1 + rowSums(txt_full_DTM_mtrx != 0))
txt_X_df[, paste0(txt_var_pfx, ".weight.sum")] <-
rowSums(txt_full_DTM_mtrx)
txt_X_df[, paste0(txt_var_pfx, ".ratio.weight.sum.wrds.n")] <-
txt_X_df[, paste0(txt_var_pfx, ".weight.sum")] /
(exp(txt_X_df[, paste0(txt_var_pfx, ".wrds.n.log")]) - 1)
txt_X_df[is.nan(txt_X_df[, paste0(txt_var_pfx, ".ratio.weight.sum.wrds.n")]),
paste0(txt_var_pfx, ".ratio.weight.sum.wrds.n")] <- 0
# Create <txt_var>.chrs.n.log
txt_X_df[, paste0(txt_var_pfx, ".chrs.n.log")] <-
log(1 + mycount_pattern_occ(".", glbObsAll[, txt_var]))
txt_X_df[, paste0(txt_var_pfx, ".chrs.uppr.n.log")] <-
log(1 + mycount_pattern_occ("[[:upper:]]", glbObsAll[, txt_var]))
txt_X_df[, paste0(txt_var_pfx, ".dgts.n.log")] <-
log(1 + mycount_pattern_occ("[[:digit:]]", glbObsAll[, txt_var]))
# Create <txt_var>.npnct?.log
# would this be faster if it's iterated over each row instead of
# each created column ???
for (punct_ix in 1:length(glb_punct_vctr)) {
# smp0 <- " "
# smp1 <- "! \" # $ % & ' ( ) * + , - . / : ; < = > ? @ [ \ ] ^ _ ` { | } ~"
# smp2 <- paste(smp1, smp1, sep=" ")
# print(sprintf("Testing %s pattern:", glb_punct_vctr[punct_ix]))
# results <- mycount_pattern_occ(glb_punct_vctr[punct_ix], c(smp0, smp1, smp2))
# names(results) <- NULL; print(results)
txt_X_df[,
paste0(txt_var_pfx, ".chrs.pnct", sprintf("%02d", punct_ix), ".n.log")] <-
log(1 + mycount_pattern_occ(glb_punct_vctr[punct_ix],
glbObsAll[, txt_var]))
}
# print(head(glbObsAll[glbObsAll[, "A.npnct23.log"] > 0,
# c("UniqueID", "Popular", "Abstract", "A.npnct23.log")]))
# Create <txt_var>.niso8859.log
txt_X_df[, paste0(txt_var_pfx, ".chrs.iso8859.n.log")] <-
log1p(mycount_pattern_occ("&#[[:digit:]]{3};", glbObsAll[, txt_var]))
# Create <txt_var>.wrds.stop.n.log & <txt_var>ratio.wrds.stop.n.wrds.n
if (!is.null(glb_txt_stop_words[[txt_var]])) {
stop_words_rex_str <- paste0("\\b(",
paste0(glb_txt_stop_words[[txt_var]], collapse="|"),
")\\b")
txt_X_df[, paste0(txt_var_pfx, ".wrds.stop.n", ".log")] <-
log(1 + mycount_pattern_occ(stop_words_rex_str, glb_txt_chr_lst[[txt_var]]))
txt_X_df[, paste0(txt_var_pfx, ".ratio.wrds.stop.n.wrds.n")] <-
exp(txt_X_df[, paste0(txt_var_pfx, ".wrds.stop.n", ".log")] -
txt_X_df[, paste0(txt_var_pfx, ".wrds.n", ".log")])
}
# Create <txt_var>.P.http
txt_X_df[, paste(txt_var_pfx, ".P.http", sep="")] <-
log1p(mycount_pattern_occ("http", glbObsAll[, txt_var]))
# Create <txt_var>.P.<user-spec-pattern>
for (pattern in names(glbFeatsTextPatterns[[txt_var]]))
txt_X_df[, paste(txt_var_pfx, ".P.", pattern, sep="")] <-
log1p(mycount_pattern_occ(glbFeatsTextPatterns[[txt_var]][pattern],
glbObsAll[, txt_var]))
txt_X_df <- subset(txt_X_df, select=-.rnorm)
txt_X_df <- txt_X_df[, -grep(glb_id_var, names(txt_X_df), fixed=TRUE), FALSE]
#glbObsAll <- cbind(glbObsAll, txt_X_df)
}
glbObsAll <- cbind(glbObsAll, txt_X_df)
#myplot_box(glbObsAll, "A.sum.TfIdf", glb_rsp_var)
# Generate summaries
# print(summary(glbObsAll))
# print(sapply(names(glbObsAll), function(col) sum(is.na(glbObsAll[, col]))))
# print(summary(glbObsTrn))
# print(sapply(names(glbObsTrn), function(col) sum(is.na(glbObsTrn[, col]))))
# print(summary(glbObsNew))
# print(sapply(names(glbObsNew), function(col) sum(is.na(glbObsNew[, col]))))
glbFeatsExclude <- union(glbFeatsExclude,
glbFeatsText)
rm(log_X_df, txt_X_df)
}
# print(sapply(names(glbObsTrn), function(col) sum(is.na(glbObsTrn[, col]))))
# print(sapply(names(glbObsNew), function(col) sum(is.na(glbObsNew[, col]))))
# print(myplot_scatter(glbObsTrn, "<col1_name>", "<col2_name>", smooth=TRUE))
#stop(here"); glb_to_sav(); glbObsAll <- sav_allobs_df
if (!is.null(glbFeatsPrice)) {
for (var in glbFeatsPrice) {
for (digit in 1:(log10(max(glbObsAll[, var], na.rm=TRUE)) + 1)) {
glbObsAll[, paste0(var, ".dgt", digit, ".is9")] <-
as.numeric(as.integer((as.integer(glbObsAll[, var]) %% (10 ^ digit)) /
(10 ^ (digit - 1))) == 9)
# glbObsAll[, paste0(var, ".dgt", digit, ".is9.fctr")] <-
# as.factor(as.integer((as.integer(glbObsAll[, var]) %% (10 ^ digit)) /
# (10 ^ (digit - 1))) == 9)
}
for (decimal in 1:2) {
glbObsAll[, paste0(var, ".dcm", decimal, ".is9")] <-
as.numeric(as.integer(glbObsAll[, var] * (10 ^ decimal)) %% 10 == 9)
# glbObsAll[, paste0(var, ".dcm", decimal, ".is9.fctr")] <-
# as.factor(as.integer(glbObsAll[, var] * (10 ^ decimal)) %% 10 == 9)
}
}
#as.numeric((as.integer(startprice) %% 10) == 9)
}
rm(corpus_lst
, glb_sprs_DTM_lst #, glb_full_DTM_lst
, txt_corpus, txt_vctr)## Warning in rm(corpus_lst, glb_sprs_DTM_lst, txt_corpus, txt_vctr): object
## 'corpus_lst' not found## Warning in rm(corpus_lst, glb_sprs_DTM_lst, txt_corpus, txt_vctr): object
## 'glb_sprs_DTM_lst' not found## Warning in rm(corpus_lst, glb_sprs_DTM_lst, txt_corpus, txt_vctr): object
## 'txt_corpus' not found## Warning in rm(corpus_lst, glb_sprs_DTM_lst, txt_corpus, txt_vctr): object
## 'txt_vctr' not foundextract.features_chunk_df <- myadd_chunk(extract.features_chunk_df, "extract.features_end",
major.inc=TRUE)## label step_major step_minor label_minor
## 2 extract.features_factorize.str.vars 2 0 0
## 3 extract.features_end 3 0 0
## bgn end elapsed
## 2 22.552 22.578 0.026
## 3 22.578 NA NAmyplt_chunk(extract.features_chunk_df)## label step_major step_minor label_minor
## 2 extract.features_factorize.str.vars 2 0 0
## 1 extract.features_bgn 1 0 0
## bgn end elapsed duration
## 2 22.552 22.578 0.026 0.026
## 1 22.539 22.552 0.013 0.013
## [1] "Total Elapsed Time: 22.578 secs"
# if (glb_save_envir)
# save(glb_feats_df,
# glbObsAll, #glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
# file=paste0(glb_out_pfx, "extract_features_dsk.RData"))
# load(paste0(glb_out_pfx, "extract_features_dsk.RData"))
replay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"data.training.all","data.new")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
#glb_chunks_df <- myadd_chunk(glb_chunks_df, "manage.missing.data", major.inc=TRUE)Step 3.0: extract features
glb_chunks_df <- myadd_chunk(glb_chunks_df, "manage.missing.data", major.inc=FALSE)## label step_major step_minor label_minor bgn end
## 5 extract.features 3 0 0 22.532 24.177
## 6 manage.missing.data 3 1 1 24.178 NA
## elapsed
## 5 1.645
## 6 NA# If mice crashes with error: Error in get(as.character(FUN), mode = "function", envir = envir) : object 'State' of mode 'function' was not found
# consider excluding 'State' as a feature
# print(sapply(names(glbObsTrn), function(col) sum(is.na(glbObsTrn[, col]))))
# print(sapply(names(glbObsNew), function(col) sum(is.na(glbObsNew[, col]))))
# glbObsTrn <- na.omit(glbObsTrn)
# glbObsNew <- na.omit(glbObsNew)
# df[is.na(df)] <- 0
mycheck_problem_data(glbObsAll, featsExclude = glbFeatsExclude,
fctrMaxUniqVals = glbFctrMaxUniqVals)## [1] "numeric data missing in glbObsAll: "
## Popular Popular.fctr
## 1870 1870
## [1] "numeric data w/ 0s in glbObsAll: "
## WordCount Popular WordCount.log1p WordCount.root2
## 109 5439 109 109
## WordCount.nexp
## 2044
## [1] "numeric data w/ Infs in glbObsAll: "
## named integer(0)
## [1] "numeric data w/ NaNs in glbObsAll: "
## named integer(0)
## [1] "string data missing in glbObsAll: "
## NewsDesk SectionName SubsectionName Headline Snippet
## 2408 2899 6176 0 13
## Abstract PubDate NDSSName.my
## 17 0 0# glbObsAll <- na.omit(glbObsAll)
# Not refactored into mydsutils.R since glb_*_df might be reassigned
glb_impute_missing_data <- function() {
require(mice)
set.seed(glb_mice_complete.seed)
inp_impent_df <- glbObsAll[, setdiff(names(glbObsAll),
union(glbFeatsExclude, glb_rsp_var))]
print("Summary before imputation: ")
print(summary(inp_impent_df))
out_impent_df <- complete(mice(inp_impent_df))
print(summary(out_impent_df))
ret_vars <- sapply(names(out_impent_df),
function(col) ifelse(!identical(out_impent_df[, col],
inp_impent_df[, col]),
col, ""))
ret_vars <- ret_vars[ret_vars != ""]
# complete(mice()) changes attributes of factors even though values don't change
for (col in ret_vars) {
if (inherits(out_impent_df[, col], "factor")) {
if (identical(as.numeric(out_impent_df[, col]),
as.numeric(inp_impent_df[, col])))
ret_vars <- setdiff(ret_vars, col)
}
}
return(out_impent_df[, ret_vars])
}
if (glb_impute_na_data &&
(length(myfind_numerics_missing(glbObsAll)) > 0) &&
(ncol(nonna_df <- glb_impute_missing_data()) > 0)) {
for (col in names(nonna_df)) {
glbObsAll[, paste0(col, ".nonNA")] <- nonna_df[, col]
glbFeatsExclude <- c(glbFeatsExclude, col)
}
}
mycheck_problem_data(glbObsAll, featsExclude = glbFeatsExclude,
fctrMaxUniqVals = glbFctrMaxUniqVals, terminate = TRUE)## [1] "numeric data missing in glbObsAll: "
## Popular Popular.fctr
## 1870 1870
## [1] "numeric data w/ 0s in glbObsAll: "
## WordCount Popular WordCount.log1p WordCount.root2
## 109 5439 109 109
## WordCount.nexp
## 2044
## [1] "numeric data w/ Infs in glbObsAll: "
## named integer(0)
## [1] "numeric data w/ NaNs in glbObsAll: "
## named integer(0)
## [1] "string data missing in glbObsAll: "
## NewsDesk SectionName SubsectionName Headline Snippet
## 2408 2899 6176 0 13
## Abstract PubDate NDSSName.my
## 17 0 0Step 3.1: manage missing data
glb_chunks_df <- myadd_chunk(glb_chunks_df, "cluster.data", major.inc=FALSE)## label step_major step_minor label_minor bgn end
## 6 manage.missing.data 3 1 1 24.178 24.259
## 7 cluster.data 3 2 2 24.259 NA
## elapsed
## 6 0.081
## 7 NAmycompute_entropy_df <- function(obs_df, entropy_var, by_var=NULL) {
require(lazyeval)
require(dplyr)
require(tidyr)
if (is.null(by_var)) {
by_var <- ".default"
obs_df$.default <- as.factor(".default")
}
if (!any(grepl(".clusterid", names(obs_df), fixed=TRUE)))
obs_df$.clusterid <- 1
cluster_df <- obs_df %>%
count_(c(by_var, ".clusterid", entropy_var)) %>%
dplyr::filter(n > 0) %>%
dplyr::filter_(interp(~(!is.na(var)), var=as.name(entropy_var))) %>%
unite_(paste0(by_var, ".clusterid"),
c(interp(by_var), ".clusterid")) %>%
spread_(interp(entropy_var), "n", fill=0)
# head(cluster_df)
# sum(cluster_df$n)
tmp.entropy <- sapply(1:nrow(cluster_df),
function(row) entropy(as.numeric(cluster_df[row, -1]), method = "ML"))
tmp.knt <- sapply(1:nrow(cluster_df),
function(row) sum(as.numeric(cluster_df[row, -1])))
cluster_df$.entropy <- tmp.entropy; cluster_df$.knt <- tmp.knt
#print(cluster_df)
return(cluster_df)
}
if (glb_cluster) {
require(proxy)
#require(hash)
require(dynamicTreeCut)
require(entropy)
require(tidyr)
require(ggdendro)
mywgtdcosine_dist <- function(x, y=NULL, weights=NULL) {
if (!inherits(x, "matrix"))
x <- as.matrix(x)
if (is.null(weights))
weights <- rep(1, ncol(x))
wgtsx <- matrix(rep(weights / sum(weights), nrow(x)), nrow = nrow(x),
byrow = TRUE)
wgtdx <- x * wgtsx
wgtdxsqsum <- as.matrix(rowSums((x ^ 2) * wgtsx), byrow=FALSE)
denom <- sqrt(wgtdxsqsum %*% t(wgtdxsqsum))
ret_mtrx <- 1 - ((sum(weights) ^ 1) * (wgtdx %*% t(wgtdx)) / denom)
ret_mtrx[is.nan(ret_mtrx)] <- 1
diag(ret_mtrx) <- 0
return(ret_mtrx)
}
#pr_DB$delete_entry("mywgtdcosine");
# Need to do this only once across runs ?
if (!pr_DB$entry_exists("mywgtdcosine")) {
pr_DB$set_entry(FUN = mywgtdcosine_dist, names = c("mywgtdcosine"))
pr_DB$modify_entry(names="mywgtdcosine", type="metric", loop=FALSE)
}
#pr_DB$get_entry("mywgtdcosine")
# glb_hash <- hash(key=unique(glbObsAll$myCategory),
# values=1:length(unique(glbObsAll$myCategory)))
# glb_hash_lst <- hash(key=unique(glbObsAll$myCategory),
# values=1:length(unique(glbObsAll$myCategory)))
#stop(here"); glb_to_sav(); glbObsAll <- sav_allobs_df
cluster_vars <- grep(paste0("[",
toupper(paste0(substr(glbFeatsText, 1, 1), collapse = "")),
"]\\.[PT]\\."),
names(glbObsAll), value = TRUE)
# Assign correlations with rsp_var as weights for cosine distance
print("Clustering features: ")
cluster_vars_df <- data.frame(abs.cor.y = abs(cor(
glbObsAll[glbObsAll$.src == "Train", cluster_vars],
as.numeric(glbObsAll[glbObsAll$.src == "Train", glb_rsp_var]),
use = "pairwise.complete.obs")))
print(tail(cluster_vars_df <- orderBy(~ abs.cor.y,
subset(cluster_vars_df, !is.na(abs.cor.y))), 5))
print(sprintf(" .rnorm cor: %0.4f",
cor(glbObsAll[glbObsAll$.src == "Train", ".rnorm"],
as.numeric(glbObsAll[glbObsAll$.src == "Train", glb_rsp_var]),
use = "pairwise.complete.obs")))
print(sprintf("glbObsAll Entropy: %0.4f",
allobs_ent <- entropy(table(glbObsAll[, glb_cluster_entropy_var]),
method="ML")))
print(category_df <- mycompute_entropy_df(obs_df=glbObsAll,
entropy_var=glb_cluster_entropy_var,
by_var=glb_category_var))
print(sprintf("glbObsAll$%s Entropy: %0.4f (%0.4f pct)",
glb_category_var,
category_ent <- weighted.mean(category_df$.entropy, category_df$.knt),
100 * category_ent / allobs_ent))
glbObsAll$.clusterid <- 1
#print(max(table(glbObsAll$myCategory.fctr) / 20))
#stop(here"); glb_to_sav()
grp_ids <- sort(unique(glbObsAll[, glb_category_var]))
glb_cluster_size_df_lst <- list()
png(paste0(glb_out_pfx, "FeatsTxtClusters.png"),
width = 480 * 2, height = 480 * length(grp_ids))
grid.newpage()
pushViewport(viewport(layout = grid.layout(nrow = length(grp_ids), ncol = 2)))
pltIx <- 1
for (grp in grp_ids) {
# if (grep(grp, levels(grp_ids)) <= 6) next
# if (grep(grp, levels(grp_ids)) > 9) next
# if (grep(grp, levels(grp_ids)) != 10) next
print(sprintf("Category: %s", grp))
ctgry_allobs_df <- glbObsAll[glbObsAll[, glb_category_var] == grp, ]
if (!inherits(ctgry_allobs_df[, glb_cluster_entropy_var], "factor"))
ctgry_allobs_df[, glb_cluster_entropy_var] <-
as.factor(ctgry_allobs_df[, glb_cluster_entropy_var])
#dstns_dist <- proxy::dist(ctgry_allobs_df[, cluster_vars], method = "cosine")
dstns_dist <- proxy::dist(ctgry_allobs_df[, row.names(cluster_vars_df)],
method = "mywgtdcosine",
weights = cluster_vars_df$abs.cor.y)
# Custom distance functions return a crossdist object
#dstns_mtrx <- as.matrix(dstns_dist)
dstns_mtrx <- matrix(as.vector(dstns_dist), nrow=attr(dstns_dist, "dim")[1],
dimnames=attr(dstns_dist, "dimnames"))
dstns_dist <- as.dist(dstns_mtrx)
print(sprintf("max distance(%0.4f) pair:", max(dstns_mtrx)))
# print(dim(dstns_mtrx))
# print(sprintf("which.max: %d", which.max(dstns_mtrx)))
row_ix <- ceiling(which.max(dstns_mtrx) / ncol(dstns_mtrx))
col_ix <- which.max(dstns_mtrx[row_ix, ])
# print(sprintf("row_ix: %d", row_ix)); print(sprintf("col_ix: %d", col_ix));
# print(dim(ctgry_allobs_df))
print(ctgry_allobs_df[c(row_ix, col_ix),
c(glb_id_var, glb_cluster_entropy_var, glb_category_var, glbFeatsText, cluster_vars)])
min_dstns_mtrx <- dstns_mtrx
diag(min_dstns_mtrx) <- 1
# Float representations issue -2.22e-16 vs. 0.0000
print(sprintf("min distance(%0.4f) pair:", min(min_dstns_mtrx)))
row_ix <- ceiling(which.min(min_dstns_mtrx) / ncol(min_dstns_mtrx))
col_ix <- which.min(min_dstns_mtrx[row_ix, ])
print(ctgry_allobs_df[c(row_ix, col_ix),
c(glb_id_var, glb_cluster_entropy_var, glb_category_var, glbFeatsText,
cluster_vars)])
set.seed(glb_cluster.seed)
clusters <- hclust(dstns_dist, method = "ward.D2")
# Workaround to avoid "Error in cutree(dendro, h = heightcutoff) : the 'height' component of 'tree' is not sorted (increasingly)"
if (with(clusters,all.equal(height,sort(height))))
clusters$height <- round(clusters$height,6)
clusters$labels <- ctgry_allobs_df[, glb_id_var]
clustersDD <- dendro_data(clusters)
clustersDD$labels[, glb_rsp_var] <- sapply(clustersDD$labels$label, function(id)
ctgry_allobs_df[id == ctgry_allobs_df[, glb_id_var], glb_rsp_var])
print(ggdendrogram(clustersDD, rotate = TRUE, size = 2) +
geom_point(data = clustersDD$labels,
aes_string(x = "x", color = glb_rsp_var), y = min(clustersDD$segments$y)) +
coord_flip(ylim = c(min(clustersDD$segments$y),
max(clustersDD$segments$y))) +
ggtitle(grp),
vp = viewport(layout.pos.row = pltIx, layout.pos.col = 1))
# clusters$labels <- ctgry_allobs_df[, glb_id_var]
# clustersDD <- dendro_data(clusters)
# clustersDD$labels$color <- sapply(clustersDD$labels$label, function(id)
# ctgry_allobs_df[id == ctgry_allobs_df[, glb_id_var], glb_rsp_var])
# print(ggdendrogram(clustersDD, rotate = TRUE, size = 2) +
# geom_point(data = clustersDD$labels,
# aes_string(x = "x", color = "color"), y = min(clustersDD$segments$y)) +
# coord_flip(ylim = c(min(clustersDD$segments$y),
# max(clustersDD$segments$y))))
# print(ggdendrogram(clustersDD, rotate = TRUE, size = 2) +
# geom_point(data = clustersDD$labels,
# aes_string(x = "x", y = "y", color = "color")))
# myplclust(clusters, lab=ctgry_allobs_df[, glb_id_var],
# lab.col=unclass(ctgry_allobs_df[, glb_cluster_entropy_var]))
opt_minclustersize_df <- data.frame(minclustersize = nrow(ctgry_allobs_df),
entropy = entropy(table(ctgry_allobs_df[, glb_cluster_entropy_var]),
method = "ML"))
for (minclustersize in
as.integer(seq(nrow(ctgry_allobs_df) / 2, nrow(ctgry_allobs_df) / 10,
length = 5))) {
clusterGroups <- cutreeDynamic(clusters, minClusterSize = minclustersize,
method = "tree", deepSplit = 0)
# Unassigned groups are labeled 0; the largest group has label 1
clusterGroups[clusterGroups == 0] <- 1
ctgry_allobs_df$.clusterid <- clusterGroups
ctgry_clstrs_df <- mycompute_entropy_df(ctgry_allobs_df,
glb_cluster_entropy_var)
opt_minclustersize_df <- rbind(opt_minclustersize_df,
data.frame(minclustersize = minclustersize,
entropy = weighted.mean(ctgry_clstrs_df$.entropy, ctgry_clstrs_df$.knt)))
}
opt_minclustersize <-
opt_minclustersize_df$minclustersize[which.min(opt_minclustersize_df$entropy)]
opt_minclustersize_df$.color <-
ifelse(opt_minclustersize_df$minclustersize == opt_minclustersize,
"red", "blue")
print(ggplot(data = opt_minclustersize_df,
mapping = aes(x = minclustersize, y = entropy)) +
geom_point(aes(color = .color)) + scale_color_identity() +
guides(color = "none") + geom_line(),
vp = viewport(layout.pos.row = pltIx, layout.pos.col = 2))
glb_cluster_size_df_lst[[grp]] <- opt_minclustersize_df
# select minclustersize that minimizes entropy
clusterGroups <- cutreeDynamic(clusters, minClusterSize = opt_minclustersize,
method = "tree",
deepSplit = 0)
# Unassigned groups are labeled 0; the largest group has label 1
table(clusterGroups, ctgry_allobs_df[, glb_cluster_entropy_var],
useNA = "ifany")
clusterGroups[clusterGroups == 0] <- 1
table(clusterGroups, ctgry_allobs_df[, glb_cluster_entropy_var], useNA = "ifany")
glbObsAll[glbObsAll[, glb_category_var] == grp,]$.clusterid <-
clusterGroups
pltIx <- pltIx + 1
}
dev.off()
#all.equal(sav_allobs_df_clusterid, glbObsAll$.clusterid)
print(cluster_df <- mycompute_entropy_df(obs_df=glbObsAll,
entropy_var=glb_cluster_entropy_var,
by_var=glb_category_var))
print(sprintf("glbObsAll$%s$.clusterid Entropy: %0.4f (%0.4f pct)",
glb_category_var,
cluster_ent <- weighted.mean(cluster_df$.entropy, cluster_df$.knt),
100 * cluster_ent / category_ent))
glbObsAll$.clusterid.fctr <- as.factor(glbObsAll$.clusterid)
# .clusterid.fctr is created automatically (probably ?) later
glbFeatsExclude <- c(glbFeatsExclude, ".clusterid")
if (!is.null(glb_category_var))
# glbFeatsInteractionOnly[ifelse(grepl("\\.fctr", glb_category_var),
# glb_category_var,
# paste0(glb_category_var, ".fctr"))] <-
# c(".clusterid.fctr")
glbFeatsInteractionOnly[[".clusterid.fctr"]] <-
ifelse(grepl("\\.fctr", glb_category_var), glb_category_var,
paste0(glb_category_var, ".fctr"))
if (glbFeatsTextClusterVarsExclude)
glbFeatsExclude <- c(glbFeatsExclude, cluster_vars)
}
# Last call for data modifications
#stop(here") # sav_allobs_df <- glbObsAll
# glbObsAll[(glbObsAll$PropR == 0.75) & (glbObsAll$State == "Hawaii"), "PropR.fctr"] <- "N"
# Re-partition
glbObsTrn <- subset(glbObsAll, .src == "Train")
glbObsNew <- subset(glbObsAll, .src == "Test")
glb_chunks_df <- myadd_chunk(glb_chunks_df, "partition.data.training", major.inc=TRUE)## label step_major step_minor label_minor bgn end
## 7 cluster.data 3 2 2 24.259 24.288
## 8 partition.data.training 4 0 0 24.288 NA
## elapsed
## 7 0.029
## 8 NAStep 4.0: partition data training
if (all(is.na(glbObsNew[, glb_rsp_var]))) {
set.seed(glb_split_sample.seed)
OOB_size <- nrow(glbObsNew) * 1.1
if (is.null(glb_category_var)) {
require(caTools)
split <- sample.split(glbObsTrn[, glb_rsp_var_raw],
SplitRatio=OOB_size / nrow(glbObsTrn))
glbObsOOB <- glbObsTrn[split ,]
glbObsFit <- glbObsTrn[!split, ]
} else {
sample_vars <- c(glb_category_var, glb_rsp_var_raw)
rspvar_freq_df <- orderBy(reformulate(glb_rsp_var_raw),
mycreate_sqlxtab_df(glbObsTrn, glb_rsp_var_raw))
OOB_rspvar_size <-
1.0 * OOB_size * rspvar_freq_df$.n / sum(rspvar_freq_df$.n)
names(OOB_rspvar_size) <- as.character(rspvar_freq_df[, glb_rsp_var_raw])
newobs_freq_df <- orderBy(reformulate(glb_category_var),
mycreate_sqlxtab_df(glbObsNew, glb_category_var))
trnobs_freq_df <- orderBy(reformulate(glb_category_var),
mycreate_sqlxtab_df(glbObsTrn, glb_category_var))
ctgry_freq_df <- merge(newobs_freq_df, trnobs_freq_df,
by = glb_category_var,
all = TRUE, sort = TRUE,
suffixes = c(".tst", ".trn"))
ctgry_freq_df[is.na(ctgry_freq_df)] <- 0
obs_freq_df <- mycreate_xtab_df(glbObsTrn,
c(glb_category_var, glb_rsp_var_raw))
newobs_freq_df <- orderBy(reformulate(glb_category_var),
mycreate_sqlxtab_df(glbObsNew, glb_category_var))
names(newobs_freq_df) <- gsub(".n", ".n.tst", names(newobs_freq_df),
fixed = TRUE)
obs_freq_df <- merge(obs_freq_df, newobs_freq_df, all = TRUE)
strata_mtrx <- ceiling(
matrix(obs_freq_df$.n.tst * 1.0 / sum(obs_freq_df$.n.tst)) %*%
matrix(OOB_rspvar_size, nrow = 1))
dimnames(strata_mtrx)[[1]] <- obs_freq_df[, glb_category_var]
dimnames(strata_mtrx)[[2]] <-
as.character(rspvar_freq_df[, glb_rsp_var_raw])
for (val in rspvar_freq_df[, glb_rsp_var_raw]) {
trn <- paste0(glb_rsp_var_raw, ".", as.character(val))
strata <- paste0(".strata.", as.character(val))
obs_freq_df[, strata] <- strata_mtrx[, as.character(val)]
if (length(ix <- which(is.na(obs_freq_df[, trn]))) > 0) {
# NA obs in a particular category
print("Prediction Hints by Catgeory:")
print(obs_freq_df[ix, ])
obs_freq_df[ix, strata] <- NA
}
if (length((ix <- which(obs_freq_df[, trn] <
2.0 * obs_freq_df[, strata]))) > 0)
# More obs in OOB compared to fit currently
obs_freq_df[ix, strata] <- floor(obs_freq_df[ix, trn] / 2.0)
if (length((ix <- which(obs_freq_df[, strata] == 0))) > 0)
obs_freq_df[ix, strata] <- 1
}
#print(colSums(obs_freq_df[, -1]))
OOB_strata_size <- as.vector(as.matrix(obs_freq_df[,
grepl("^\\.strata\\.", names(obs_freq_df))]))
tmp_trnobs_df <- orderBy(reformulate(c(glb_rsp_var_raw, glb_category_var)),
glbObsTrn)
require(sampling)
split_strata <- sampling::strata(tmp_trnobs_df,
stratanames = c(glb_rsp_var_raw, glb_category_var),
size = OOB_strata_size[!is.na(OOB_strata_size)],
method = "srswor")
glbObsOOB <- getdata(tmp_trnobs_df, split_strata)[, names(glbObsTrn)]
glbObsFit <- glbObsTrn[!glbObsTrn[, glb_id_var] %in%
glbObsOOB[, glb_id_var], ]
}
} else {
print(sprintf("Newdata contains non-NA data for %s; setting OOB to Newdata",
glb_rsp_var))
glbObsFit <- glbObsTrn; glbObsOOB <- glbObsNew
}## [1] "Prediction Hints by Catgeory:"
## NDSSName.my.fctr Popular.0 Popular.1 .n.tst .strata.0 .strata.1
## 5 #U.S.#Education 325 NA 89 82 17
## 10 Culture## 1 NA 70 1 13
## 12 Foreign#World# 172 NA 47 44 9
## 21 myOther 38 NA 5 5 1## Loading required package: sampling
##
## Attaching package: 'sampling'
##
## The following objects are masked from 'package:survival':
##
## cluster, strata
##
## The following object is masked from 'package:caret':
##
## clusterif (!is.null(glb_max_fitobs) && (nrow(glbObsFit) > glb_max_fitobs)) {
warning("glbObsFit restricted to glb_max_fitobs: ",
format(glb_max_fitobs, big.mark = ","))
org_fitobs_df <- glbObsFit
glbObsFit <-
org_fitobs_df[split <- sample.split(org_fitobs_df[, glb_rsp_var_raw],
SplitRatio = glb_max_fitobs), ]
org_fitobs_df <- NULL
}
glbObsAll$.lcn <- ""; glbObsTrn$.lcn <- "";
glbObsAll[glbObsAll[, glb_id_var] %in%
glbObsFit[, glb_id_var], ".lcn"] <- "Fit"
glbObsTrn[glbObsTrn[, glb_id_var] %in%
glbObsFit[, glb_id_var], ".lcn"] <- "Fit"
glbObsAll[glbObsAll[, glb_id_var] %in%
glbObsOOB[, glb_id_var], ".lcn"] <- "OOB"
glbObsTrn[glbObsTrn[, glb_id_var] %in%
glbObsOOB[, glb_id_var], ".lcn"] <- "OOB"
dsp_class_dstrb <- function(obs_df, location_var, partition_var) {
xtab_df <- mycreate_xtab_df(obs_df, c(location_var, partition_var))
rownames(xtab_df) <- xtab_df[, location_var]
xtab_df <- xtab_df[, -grepl(location_var, names(xtab_df))]
print(xtab_df)
print(xtab_df / rowSums(xtab_df, na.rm=TRUE))
}
# Ensure proper splits by glb_rsp_var_raw & user-specified feature for OOB vs. new
if (!is.null(glb_category_var)) {
if (glb_is_classification)
dsp_class_dstrb(glbObsAll, ".lcn", glb_rsp_var_raw)
newobs_ctgry_df <- mycreate_sqlxtab_df(subset(glbObsAll, .src == "Test"),
glb_category_var)
names(newobs_ctgry_df) <-
gsub(".n", ".n.Tst", names(newobs_ctgry_df), fixed = TRUE)
OOBobs_ctgry_df <- mycreate_sqlxtab_df(subset(glbObsAll, .lcn == "OOB"),
glb_category_var)
names(OOBobs_ctgry_df) <-
gsub(".n", ".n.OOB", names(OOBobs_ctgry_df), fixed = TRUE)
fitobs_ctgry_df <- mycreate_sqlxtab_df(subset(glbObsAll, .lcn == "Fit"),
glb_category_var)
names(fitobs_ctgry_df) <-
gsub(".n", ".n.Fit", names(fitobs_ctgry_df), fixed = TRUE)
# glb_ctgry_df <- merge(newobs_ctgry_df, OOBobs_ctgry_df, by=glb_category_var
# , all=TRUE, suffixes=c(".Tst", ".OOB"))
glb_ctgry_df <- merge(fitobs_ctgry_df, OOBobs_ctgry_df, by = glb_category_var
, all = TRUE)
glb_ctgry_df <- merge(glb_ctgry_df, newobs_ctgry_df, by = glb_category_var
, all = TRUE)
glb_ctgry_df$.freqRatio.Fit <-
glb_ctgry_df$.n.Fit / sum(glb_ctgry_df$.n.Fit, na.rm = TRUE)
glb_ctgry_df$.freqRatio.OOB <-
glb_ctgry_df$.n.OOB / sum(glb_ctgry_df$.n.OOB, na.rm = TRUE)
glb_ctgry_df$.freqRatio.Tst <-
glb_ctgry_df$.n.Tst / sum(glb_ctgry_df$.n.Tst, na.rm = TRUE)
print(orderBy(~-.freqRatio.Tst-.freqRatio.OOB-.freqRatio.Fit, glb_ctgry_df))
}## Popular.0 Popular.1 Popular.NA
## NA NA 1870
## Fit 3941 863 NA
## OOB 1498 230 NA
## Popular.0 Popular.1 Popular.NA
## NA NA 1
## Fit 0.8203580 0.1796420 NA
## OOB 0.8668981 0.1331019 NA
## NDSSName.my.fctr .n.Fit .n.OOB .n.Tst .freqRatio.Fit
## 1 ## 913 371 342 0.190049958
## 6 Business#BusinessDay#Dealbook 629 323 304 0.130932556
## 11 Culture#Arts# 490 185 174 0.101998335
## 15 OpEd#Opinion# 437 89 164 0.090965862
## 9 Business#Technology# 213 126 114 0.044338052
## 19 TStyle## 623 101 105 0.129683597
## 5 #U.S.#Education 243 82 89 0.050582848
## 10 Culture## NA 1 70 NA
## 14 Metro#N.Y./Region# 128 70 67 0.026644463
## 18 Styles#U.S.# 127 50 61 0.026436303
## 16 Science#Health# 148 48 57 0.030807660
## 13 Foreign#World#AsiaPacific 150 53 56 0.031223980
## 2 #Multimedia# 92 49 52 0.019150708
## 12 Foreign#World# 128 44 47 0.026644463
## 8 Business#Crosswords/Games# 105 18 42 0.021856786
## 7 Business#BusinessDay#SmallBusiness 100 40 41 0.020815987
## 20 Travel#Travel# 83 34 35 0.017277269
## 3 #Opinion#RoomForDebate 42 20 20 0.008742714
## 17 Styles##Fashion 104 15 15 0.021648626
## 4 #Opinion#ThePublicEditor 16 4 10 0.003330558
## 21 myOther 33 5 5 0.006869276
## .freqRatio.OOB .freqRatio.Tst
## 1 0.2146990741 0.182887701
## 6 0.1869212963 0.162566845
## 11 0.1070601852 0.093048128
## 15 0.0515046296 0.087700535
## 9 0.0729166667 0.060962567
## 19 0.0584490741 0.056149733
## 5 0.0474537037 0.047593583
## 10 0.0005787037 0.037433155
## 14 0.0405092593 0.035828877
## 18 0.0289351852 0.032620321
## 16 0.0277777778 0.030481283
## 13 0.0306712963 0.029946524
## 2 0.0283564815 0.027807487
## 12 0.0254629630 0.025133690
## 8 0.0104166667 0.022459893
## 7 0.0231481481 0.021925134
## 20 0.0196759259 0.018716578
## 3 0.0115740741 0.010695187
## 17 0.0086805556 0.008021390
## 4 0.0023148148 0.005347594
## 21 0.0028935185 0.002673797print("glbObsAll: "); print(dim(glbObsAll))## [1] "glbObsAll: "## [1] 8402 19print("glbObsTrn: "); print(dim(glbObsTrn))## [1] "glbObsTrn: "## [1] 6532 19print("glbObsFit: "); print(dim(glbObsFit))## [1] "glbObsFit: "## [1] 4804 18print("glbObsOOB: "); print(dim(glbObsOOB))## [1] "glbObsOOB: "## [1] 1728 18print("glbObsNew: "); print(dim(glbObsNew))## [1] "glbObsNew: "## [1] 1870 18# # Does not handle NULL or length(glb_id_var) > 1
if (glb_save_envir)
save(glbObsAll, #glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
file=paste0(glb_out_pfx, "blddfs_dsk.RData"))
# load(paste0(glb_out_pfx, "blddfs_dsk.RData"))
rm(split)## Warning in rm(split): object 'split' not foundglb_chunks_df <- myadd_chunk(glb_chunks_df, "select.features", major.inc=TRUE)## label step_major step_minor label_minor bgn end
## 8 partition.data.training 4 0 0 24.288 25.428
## 9 select.features 5 0 0 25.428 NA
## elapsed
## 8 1.14
## 9 NAStep 5.0: select features
#stop(here"); glb_to_sav(); glbObsAll <- sav_allobs_df
print(glb_feats_df <- myselect_features(entity_df=glbObsTrn,
exclude_vars_as_features=glbFeatsExclude,
rsp_var=glb_rsp_var))## id cor.y exclude.as.feat cor.y.abs
## Popular Popular 1.000000000 1 1.000000000
## WordCount.root2 WordCount.root2 0.292120679 0 0.292120679
## WordCount WordCount 0.257526549 1 0.257526549
## WordCount.log1p WordCount.log1p 0.254319628 0 0.254319628
## NDSSName.my.fctr NDSSName.my.fctr 0.165445970 0 0.165445970
## WordCount.nexp WordCount.nexp -0.053208396 0 0.053208396
## UniqueID UniqueID 0.011824920 1 0.011824920
## .rnorm .rnorm 0.008212201 0 0.008212201print(glb_feats_df <- orderBy(~-cor.y,
myfind_cor_features(feats_df=glb_feats_df, obs_df=glbObsTrn, rsp_var=glb_rsp_var,
nzv.freqCut=glb_nzv_freqCut, nzv.uniqueCut=glb_nzv_uniqueCut)))## [1] "cor(WordCount.log1p, WordCount.root2)=0.8906"
## [1] "cor(Popular.fctr, WordCount.log1p)=0.2543"
## [1] "cor(Popular.fctr, WordCount.root2)=0.2921"## Warning in myfind_cor_features(feats_df = glb_feats_df, obs_df =
## glbObsTrn, : Identified WordCount.log1p as highly correlated with
## WordCount.root2## id cor.y exclude.as.feat cor.y.abs
## Popular Popular 1.000000000 1 1.000000000
## WordCount.root2 WordCount.root2 0.292120679 0 0.292120679
## WordCount WordCount 0.257526549 1 0.257526549
## WordCount.log1p WordCount.log1p 0.254319628 0 0.254319628
## NDSSName.my.fctr NDSSName.my.fctr 0.165445970 0 0.165445970
## UniqueID UniqueID 0.011824920 1 0.011824920
## .rnorm .rnorm 0.008212201 0 0.008212201
## WordCount.nexp WordCount.nexp -0.053208396 0 0.053208396
## cor.high.X freqRatio percentUnique zeroVar nzv
## Popular <NA> 4.976212 0.03061849 FALSE FALSE
## WordCount.root2 <NA> 2.315789 24.15799143 FALSE FALSE
## WordCount <NA> 2.315789 24.15799143 FALSE FALSE
## WordCount.log1p WordCount.root2 2.315789 24.15799143 FALSE FALSE
## NDSSName.my.fctr <NA> 1.348739 0.32149418 FALSE FALSE
## UniqueID <NA> 1.000000 100.00000000 FALSE FALSE
## .rnorm <NA> 1.000000 100.00000000 FALSE FALSE
## WordCount.nexp <NA> 17.761364 11.32884262 FALSE FALSE
## is.cor.y.abs.low
## Popular FALSE
## WordCount.root2 FALSE
## WordCount FALSE
## WordCount.log1p FALSE
## NDSSName.my.fctr FALSE
## UniqueID FALSE
## .rnorm FALSE
## WordCount.nexp FALSEplt_feats_df <- glb_feats_df
print(myplot_scatter(plt_feats_df, "percentUnique", "freqRatio",
colorcol_name="nzv", jitter=TRUE) +
#geom_point(aes(shape=nzv)) +
geom_point() +
xlim(-5, 25) +
geom_hline(yintercept=glb_nzv_freqCut) +
geom_vline(xintercept=glb_nzv_uniqueCut))## Warning in myplot_scatter(plt_feats_df, "percentUnique", "freqRatio",
## colorcol_name = "nzv", : converting nzv to class:factor## Warning: Removed 2 rows containing missing values (geom_point).## Warning: Removed 2 rows containing missing values (geom_point).## Warning: Removed 2 rows containing missing values (geom_point).
print(subset(glb_feats_df, nzv))## [1] id cor.y exclude.as.feat cor.y.abs
## [5] cor.high.X freqRatio percentUnique zeroVar
## [9] nzv is.cor.y.abs.low
## <0 rows> (or 0-length row.names)tmp_allobs_df <-
glbObsAll[, union(setdiff(names(glbObsAll), subset(glb_feats_df, nzv)$id),
glb_cluster_entropy_var)]
glbObsTrn <- subset(tmp_allobs_df, .src == "Train")
glbObsNew <- subset(tmp_allobs_df, .src == "Test")
glb_feats_df$interaction.feat <- NA
for (feat in names(glbFeatsInteractionOnly))
glb_feats_df[glb_feats_df$id %in% feat, "interaction.feat"] <-
glbFeatsInteractionOnly[[feat]]
#stop(here"); glb_to_sav(); glbObsAll <- sav_allobs_df
indep_vars <- subset(glb_feats_df, !nzv & (exclude.as.feat != 1))[, "id"]
numeric_indep_vars <- indep_vars[!grepl(".fctr", indep_vars, fixed=TRUE)]
glb_feats_df$shapiro.test.p.value <- NA
glb_feats_df[glb_feats_df$id %in% numeric_indep_vars, "shapiro.test.p.value"] <-
sapply(numeric_indep_vars, function(var)
shapiro.test(glbObsTrn[, var][1:min(5000, nrow(glbObsTrn))])$p.value)
not_nrml_feats_df <- glb_feats_df %>%
subset(!is.na(shapiro.test.p.value)) %>%
subset((shapiro.test.p.value < 0.05) || (id == ".rnorm")) %>%
arrange(shapiro.test.p.value)
row.names(not_nrml_feats_df) <- not_nrml_feats_df$id
#pltTrnObs <- glbObsTrn[, c("D.npnct05.log", ".rnorm")]
pltTrnObs <- glbObsTrn[, c(union(".rnorm",
not_nrml_feats_df$id[1:min(5, nrow(not_nrml_feats_df))]),
glb_cluster_entropy_var)]
gp <- myplot_violin(pltTrnObs,
setdiff(names(pltTrnObs), glb_cluster_entropy_var),
xcol_name = glb_cluster_entropy_var)
if ("variable" %in% names(pltTrnObs))
gp <- gp + facet_wrap(~variable, scales = "free")
print(gp)
#myplot_histogram(pltTrnObs, "D.npnct11.log", fill_col_name="sold", show_stats = TRUE)
myadjust_interaction_feats <- function(vars_vctr) {
for (feat in subset(glb_feats_df, !is.na(interaction.feat))$id)
if (feat %in% vars_vctr)
vars_vctr <- union(setdiff(vars_vctr, feat),
paste0(glb_feats_df[glb_feats_df$id == feat, "interaction.feat"], ":",
feat))
return(vars_vctr)
}
myrun_rfe <- function(obs_df, indep_vars, sizes = NULL) {
rfe_obs_df <- myget_vectorized_obs_df(obs_df, glb_rsp_var, indep_vars)
predictors_vctr <- setdiff(names(rfe_obs_df), glb_rsp_var)
if (glb_is_regression) rfeFuncs <- lmFuncs else {
rfeFuncs <- ldaFuncs
# Delete non-variant columns
predictors_unqLen <- sapply(predictors_vctr, function(col)
length(unique(rfe_obs_df[, col])))
predictors_vctr <- predictors_vctr[predictors_unqLen > 1]
# Delete freqRatio >= 291
# plagiarized from caret:::nzv
predictors_freqRatio <-
apply(rfe_obs_df[, predictors_vctr, FALSE], 2, function(data) {
t <- table(data[!is.na(data)])
if (length(t) <= 1) {
return(0)
}
w <- which.max(t)
return(max(t, na.rm = TRUE)/max(t[-w], na.rm = TRUE))
})
predictors_vctr <- predictors_vctr[predictors_freqRatio < 172]
}
if (is.null(sizes))
sizes <- tail(2 ^ (1:as.integer(log2(length(predictors_vctr)))), 5)
rfe_control <- rfeControl(functions = rfeFuncs, method = "repeatedcv",
number = glb_rcv_n_folds, repeats = glb_rcv_n_repeats,
verbose = TRUE, returnResamp = "all",
seeds = mygen_seeds(seeds_lst_len =
(glb_rcv_n_folds * glb_rcv_n_repeats) + 1,
seeds_elmnt_lst_len = (length(sizes) + 1))
, allowParallel = FALSE
)
set.seed(113)
rfe_results <- rfe(rfe_obs_df[, predictors_vctr, FALSE],
rfe_obs_df[, glb_rsp_var],
sizes = sizes,
# metric = unlist(strsplit(glbMdlMetricsEval, "[.]"))[2],
# maximize = ifelse(unlist(strsplit(glbMdlMetricsEval, "[.]"))[1] == "max",
# TRUE, FALSE),
rfeControl = rfe_control)
print(rfe_results)
print(predictors(rfe_results))
# print(plot(rfe_results, type=c("g", "o")))
# print(plot(rfe_results))
print(ggplot(rfe_results))
return(rfe_results)
}
#stop(here"); glb_to_sav()
# shd .clusterid.fctr be excluded from this ? or include encoding of glb_category_var:.clusterid.fctr ?
indep_vars <- myadjust_interaction_feats(subset(glb_feats_df,
!nzv & (exclude.as.feat != 1))$id)
rfe_fit_results <- myrun_rfe(obs_df = glbObsFit, indep_vars = indep_vars,
sizes = glbRFESizes[["RFE.X"]])## +(rfe) fit Fold1.Rep1 size: 23
## -(rfe) fit Fold1.Rep1 size: 23
## +(rfe) imp Fold1.Rep1
## -(rfe) imp Fold1.Rep1
## +(rfe) fit Fold1.Rep1 size: 16
## -(rfe) fit Fold1.Rep1 size: 16
## +(rfe) fit Fold1.Rep1 size: 8
## -(rfe) fit Fold1.Rep1 size: 8
## +(rfe) fit Fold1.Rep1 size: 4
## -(rfe) fit Fold1.Rep1 size: 4
## +(rfe) fit Fold1.Rep1 size: 2
## -(rfe) fit Fold1.Rep1 size: 2
## +(rfe) fit Fold2.Rep1 size: 23
## -(rfe) fit Fold2.Rep1 size: 23
## +(rfe) imp Fold2.Rep1
## -(rfe) imp Fold2.Rep1
## +(rfe) fit Fold2.Rep1 size: 16
## -(rfe) fit Fold2.Rep1 size: 16
## +(rfe) fit Fold2.Rep1 size: 8
## -(rfe) fit Fold2.Rep1 size: 8
## +(rfe) fit Fold2.Rep1 size: 4
## -(rfe) fit Fold2.Rep1 size: 4
## +(rfe) fit Fold2.Rep1 size: 2
## -(rfe) fit Fold2.Rep1 size: 2
## +(rfe) fit Fold3.Rep1 size: 23
## -(rfe) fit Fold3.Rep1 size: 23
## +(rfe) imp Fold3.Rep1
## -(rfe) imp Fold3.Rep1
## +(rfe) fit Fold3.Rep1 size: 16
## -(rfe) fit Fold3.Rep1 size: 16
## +(rfe) fit Fold3.Rep1 size: 8
## -(rfe) fit Fold3.Rep1 size: 8
## +(rfe) fit Fold3.Rep1 size: 4
## -(rfe) fit Fold3.Rep1 size: 4
## +(rfe) fit Fold3.Rep1 size: 2
## -(rfe) fit Fold3.Rep1 size: 2
## +(rfe) fit Fold1.Rep2 size: 23
## -(rfe) fit Fold1.Rep2 size: 23
## +(rfe) imp Fold1.Rep2
## -(rfe) imp Fold1.Rep2
## +(rfe) fit Fold1.Rep2 size: 16
## -(rfe) fit Fold1.Rep2 size: 16
## +(rfe) fit Fold1.Rep2 size: 8
## -(rfe) fit Fold1.Rep2 size: 8
## +(rfe) fit Fold1.Rep2 size: 4
## -(rfe) fit Fold1.Rep2 size: 4
## +(rfe) fit Fold1.Rep2 size: 2
## -(rfe) fit Fold1.Rep2 size: 2
## +(rfe) fit Fold2.Rep2 size: 23
## -(rfe) fit Fold2.Rep2 size: 23
## +(rfe) imp Fold2.Rep2
## -(rfe) imp Fold2.Rep2
## +(rfe) fit Fold2.Rep2 size: 16
## -(rfe) fit Fold2.Rep2 size: 16
## +(rfe) fit Fold2.Rep2 size: 8
## -(rfe) fit Fold2.Rep2 size: 8
## +(rfe) fit Fold2.Rep2 size: 4
## -(rfe) fit Fold2.Rep2 size: 4
## +(rfe) fit Fold2.Rep2 size: 2
## -(rfe) fit Fold2.Rep2 size: 2
## +(rfe) fit Fold3.Rep2 size: 23
## -(rfe) fit Fold3.Rep2 size: 23
## +(rfe) imp Fold3.Rep2
## -(rfe) imp Fold3.Rep2
## +(rfe) fit Fold3.Rep2 size: 16
## -(rfe) fit Fold3.Rep2 size: 16
## +(rfe) fit Fold3.Rep2 size: 8
## -(rfe) fit Fold3.Rep2 size: 8
## +(rfe) fit Fold3.Rep2 size: 4
## -(rfe) fit Fold3.Rep2 size: 4
## +(rfe) fit Fold3.Rep2 size: 2
## -(rfe) fit Fold3.Rep2 size: 2
## +(rfe) fit Fold1.Rep3 size: 23
## -(rfe) fit Fold1.Rep3 size: 23
## +(rfe) imp Fold1.Rep3
## -(rfe) imp Fold1.Rep3
## +(rfe) fit Fold1.Rep3 size: 16
## -(rfe) fit Fold1.Rep3 size: 16
## +(rfe) fit Fold1.Rep3 size: 8
## -(rfe) fit Fold1.Rep3 size: 8
## +(rfe) fit Fold1.Rep3 size: 4
## -(rfe) fit Fold1.Rep3 size: 4
## +(rfe) fit Fold1.Rep3 size: 2
## -(rfe) fit Fold1.Rep3 size: 2
## +(rfe) fit Fold2.Rep3 size: 23
## -(rfe) fit Fold2.Rep3 size: 23
## +(rfe) imp Fold2.Rep3
## -(rfe) imp Fold2.Rep3
## +(rfe) fit Fold2.Rep3 size: 16
## -(rfe) fit Fold2.Rep3 size: 16
## +(rfe) fit Fold2.Rep3 size: 8
## -(rfe) fit Fold2.Rep3 size: 8
## +(rfe) fit Fold2.Rep3 size: 4
## -(rfe) fit Fold2.Rep3 size: 4
## +(rfe) fit Fold2.Rep3 size: 2
## -(rfe) fit Fold2.Rep3 size: 2
## +(rfe) fit Fold3.Rep3 size: 23
## -(rfe) fit Fold3.Rep3 size: 23
## +(rfe) imp Fold3.Rep3
## -(rfe) imp Fold3.Rep3
## +(rfe) fit Fold3.Rep3 size: 16
## -(rfe) fit Fold3.Rep3 size: 16
## +(rfe) fit Fold3.Rep3 size: 8
## -(rfe) fit Fold3.Rep3 size: 8
## +(rfe) fit Fold3.Rep3 size: 4
## -(rfe) fit Fold3.Rep3 size: 4
## +(rfe) fit Fold3.Rep3 size: 2
## -(rfe) fit Fold3.Rep3 size: 2
##
## Recursive feature selection
##
## Outer resampling method: Cross-Validated (3 fold, repeated 3 times)
##
## Resampling performance over subset size:
##
## Variables Accuracy Kappa AccuracySD KappaSD Selected
## 2 0.8096 0.03708 0.005112 0.01788
## 4 0.8864 0.52518 0.005536 0.02625
## 8 0.9304 0.75904 0.004609 0.01699
## 16 0.9301 0.75823 0.004601 0.01698
## 23 0.9325 0.76819 0.004975 0.01772 *
##
## The top 5 variables (out of 23):
## WordCount.log1p, WordCount.root2, WordCount.nexp, NDSSName.my.fctrOpEd#Opinion#, NDSSName.my.fctrScience#Health#
##
## [1] "WordCount.log1p"
## [2] "WordCount.root2"
## [3] "WordCount.nexp"
## [4] "NDSSName.my.fctrOpEd#Opinion#"
## [5] "NDSSName.my.fctrScience#Health#"
## [6] "NDSSName.my.fctrBusiness#Crosswords/Games#"
## [7] "NDSSName.my.fctrStyles#U.S.#"
## [8] ".rnorm"
## [9] "NDSSName.my.fctrmyOther"
## [10] "NDSSName.my.fctr#Opinion#RoomForDebate"
## [11] "NDSSName.my.fctrBusiness#Technology#"
## [12] "NDSSName.my.fctrMetro#N.Y./Region#"
## [13] "NDSSName.my.fctrTravel#Travel#"
## [14] "NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness"
## [15] "NDSSName.my.fctr#Multimedia#"
## [16] "NDSSName.my.fctrStyles##Fashion"
## [17] "NDSSName.my.fctrForeign#World#"
## [18] "NDSSName.my.fctrForeign#World#AsiaPacific"
## [19] "NDSSName.my.fctr#U.S.#Education"
## [20] "NDSSName.my.fctrCulture#Arts#"
## [21] "NDSSName.my.fctrBusiness#BusinessDay#Dealbook"
## [22] "NDSSName.my.fctr##"
## [23] "NDSSName.my.fctrTStyle##"
# print(all.equal(rfe_results[-which(names(rfe_results) == "times")],
# sav_rfe_results[-which(names(sav_rfe_results) == "times")]))
# require(mRMRe)
# indep_vars_vctr <- subset(glb_feats_df, !nzv &
# (exclude.as.feat != 1))[, "id"]
# indep_vars_vctr <- setdiff(indep_vars_vctr,
# myfind_fctr_cols_df(glbObsTrn[, c(glb_rsp_var, indep_vars_vctr)]))
# tmp_trnobs_df <- glbObsTrn[, c(glb_rsp_var, indep_vars_vctr)]
# tmp_trnobs_df$biddable <- as.numeric(tmp_trnobs_df$biddable)
# dd <- mRMR.data(data = tmp_trnobs_df)
# mRMRe.fltr <- mRMR.classic(data = dd, target_indices = c(1), feature_count = 10)
# print(solutions(mRMRe.fltr)[[1]])
# print(apply(solutions(mRMRe.fltr)[[1]], 2, function(x, y) { return(y[x]) },
# y=featureNames(dd)))
# print(featureNames(dd)[solutions(mRMRe.fltr)[[1]]])
# print(mRMRe.fltr@filters); print(mRMRe.fltr@scores)
mycheck_problem_data(glbObsAll, featsExclude = glbFeatsExclude,
fctrMaxUniqVals = glbFctrMaxUniqVals, terminate = TRUE)## [1] "numeric data missing in glbObsAll: "
## Popular Popular.fctr
## 1870 1870
## [1] "numeric data w/ 0s in glbObsAll: "
## WordCount Popular WordCount.log1p WordCount.root2
## 109 5439 109 109
## WordCount.nexp
## 2044
## [1] "numeric data w/ Infs in glbObsAll: "
## named integer(0)
## [1] "numeric data w/ NaNs in glbObsAll: "
## named integer(0)
## [1] "string data missing in glbObsAll: "
## NewsDesk SectionName SubsectionName Headline Snippet
## 2408 2899 6176 0 13
## Abstract PubDate NDSSName.my .lcn
## 17 0 0 1870# glbObsAll %>% filter(is.na(Married.fctr)) %>% tbl_df()
# glbObsAll %>% count(Married.fctr)
# levels(glbObsAll$Married.fctr)
print("glb_feats_df:"); print(dim(glb_feats_df))## [1] "glb_feats_df:"## [1] 8 12sav_feats_df <- glb_feats_df
glb_feats_df <- sav_feats_df
glb_feats_df[, "rsp_var_raw"] <- FALSE
glb_feats_df[glb_feats_df$id == glb_rsp_var_raw, "rsp_var_raw"] <- TRUE
glb_feats_df$exclude.as.feat <- (glb_feats_df$exclude.as.feat == 1)
if (!is.null(glb_id_var) && glb_id_var != ".rownames")
glb_feats_df[glb_feats_df$id %in% glb_id_var, "id_var"] <- TRUE
add_feats_df <- data.frame(id=glb_rsp_var, exclude.as.feat=TRUE, rsp_var=TRUE)
row.names(add_feats_df) <- add_feats_df$id; print(add_feats_df)## id exclude.as.feat rsp_var
## Popular.fctr Popular.fctr TRUE TRUEglb_feats_df <- myrbind_df(glb_feats_df, add_feats_df)
if (glb_id_var != ".rownames")
print(subset(glb_feats_df, rsp_var_raw | rsp_var | id_var)) else
print(subset(glb_feats_df, rsp_var_raw | rsp_var)) ## id cor.y exclude.as.feat cor.y.abs cor.high.X
## Popular Popular 1.00000000 TRUE 1.00000000 <NA>
## UniqueID UniqueID 0.01182492 TRUE 0.01182492 <NA>
## Popular.fctr Popular.fctr NA TRUE NA <NA>
## freqRatio percentUnique zeroVar nzv is.cor.y.abs.low
## Popular 4.976212 0.03061849 FALSE FALSE FALSE
## UniqueID 1.000000 100.00000000 FALSE FALSE FALSE
## Popular.fctr NA NA NA NA NA
## interaction.feat shapiro.test.p.value rsp_var_raw id_var
## Popular NA NA TRUE NA
## UniqueID NA NA FALSE TRUE
## Popular.fctr NA NA NA NA
## rsp_var
## Popular NA
## UniqueID NA
## Popular.fctr TRUEprint("glb_feats_df vs. glbObsAll: "); ## [1] "glb_feats_df vs. glbObsAll: "print(setdiff(glb_feats_df$id, names(glbObsAll)))## character(0)print("glbObsAll vs. glb_feats_df: "); ## [1] "glbObsAll vs. glb_feats_df: "# Ensure these are only chr vars
print(setdiff(setdiff(names(glbObsAll), glb_feats_df$id),
myfind_chr_cols_df(glbObsAll)))## character(0)if (glb_save_envir)
save(glb_feats_df,
glbObsAll, #glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
file=paste0(glb_out_pfx, "selfts_dsk.RData"))
# load(paste0(glb_out_pfx, "blddfs_dsk.RData"))
# if (!all.equal(tmp_feats_df, glb_feats_df))
# stop("glb_feats_df r/w not working")
# if (!all.equal(tmp_entity_df, glbObsAll))
# stop("glbObsAll r/w not working")
glb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.models", major.inc=TRUE)## label step_major step_minor label_minor bgn end elapsed
## 9 select.features 5 0 0 25.428 35.237 9.809
## 10 fit.models 6 0 0 35.238 NA NAStep 6.0: fit models
# load(paste0(glb_out_pfx, "dsk.RData"))
get_model_sel_frmla <- function() {
model_evl_terms <- c(NULL)
# min.aic.fit might not be avl
lclMdlEvlCriteria <-
glbMdlMetricsEval[glbMdlMetricsEval %in% names(glb_models_df)]
for (metric in lclMdlEvlCriteria)
model_evl_terms <- c(model_evl_terms,
ifelse(length(grep("max", metric)) > 0, "-", "+"), metric)
if (glb_is_classification && glb_is_binomial)
model_evl_terms <- c(model_evl_terms, "-", "opt.prob.threshold.OOB")
model_sel_frmla <- as.formula(paste(c("~ ", model_evl_terms), collapse = " "))
return(model_sel_frmla)
}
get_dsp_models_df <- function() {
dsp_models_cols <- c("id",
glbMdlMetricsEval[glbMdlMetricsEval %in% names(glb_models_df)],
grep("opt.", names(glb_models_df), fixed = TRUE, value = TRUE))
dsp_models_df <-
#orderBy(get_model_sel_frmla(), glb_models_df)[, c("id", glbMdlMetricsEval)]
orderBy(get_model_sel_frmla(), glb_models_df)[, dsp_models_cols]
nCvMdl <- sapply(glb_models_lst, function(mdl) nrow(mdl$results))
nParams <- sapply(glb_models_lst, function(mdl) ifelse(mdl$method == "custom", 0,
nrow(subset(modelLookup(mdl$method), parameter != "parameter"))))
# nCvMdl <- nCvMdl[names(nCvMdl) != "avNNet"]
# nParams <- nParams[names(nParams) != "avNNet"]
if (length(cvMdlProblems <- nCvMdl[nCvMdl <= nParams]) > 0) {
print("Cross Validation issues:")
warning("Cross Validation issues:")
print(cvMdlProblems)
}
pltMdls <- setdiff(names(nCvMdl), names(cvMdlProblems))
pltMdls <- setdiff(pltMdls, names(nParams[nParams == 0]))
# length(pltMdls) == 21
png(paste0(glb_out_pfx, "bestTune.png"), width = 480 * 2, height = 480 * 4)
grid.newpage()
pushViewport(viewport(layout = grid.layout(ceiling(length(pltMdls) / 2.0), 2)))
pltIx <- 1
for (mdlId in pltMdls) {
print(ggplot(glb_models_lst[[mdlId]], highBestTune = TRUE) + labs(title = mdlId),
vp = viewport(layout.pos.row = ceiling(pltIx / 2.0),
layout.pos.col = ((pltIx - 1) %% 2) + 1))
pltIx <- pltIx + 1
}
dev.off()
return(dsp_models_df)
}
#get_dsp_models_df()
if (glb_is_classification && glb_is_binomial &&
(length(unique(glbObsFit[, glb_rsp_var])) < 2))
stop("glbObsFit$", glb_rsp_var, ": contains less than 2 unique values: ",
paste0(unique(glbObsFit[, glb_rsp_var]), collapse=", "))
max_cor_y_x_vars <- orderBy(~ -cor.y.abs,
subset(glb_feats_df, (exclude.as.feat == 0) & !nzv & !is.cor.y.abs.low &
is.na(cor.high.X)))[1:2, "id"]
max_cor_y_x_vars <- max_cor_y_x_vars[!is.na(max_cor_y_x_vars)]
if (!is.null(glb_Baseline_mdl_var)) {
if ((max_cor_y_x_vars[1] != glb_Baseline_mdl_var) &
(glb_feats_df[glb_feats_df$id == max_cor_y_x_vars[1], "cor.y.abs"] >
glb_feats_df[glb_feats_df$id == glb_Baseline_mdl_var, "cor.y.abs"]))
stop(max_cor_y_x_vars[1], " has a higher correlation with ", glb_rsp_var,
" than the Baseline var: ", glb_Baseline_mdl_var)
}
glb_model_type <- ifelse(glb_is_regression, "regression", "classification")
# Model specs
c("id.prefix", "method", "type",
# trainControl params
"preProc.method", "cv.n.folds", "cv.n.repeats", "summary.fn",
# train params
"metric", "metric.maximize", "tune.df")## [1] "id.prefix" "method" "type"
## [4] "preProc.method" "cv.n.folds" "cv.n.repeats"
## [7] "summary.fn" "metric" "metric.maximize"
## [10] "tune.df"# Baseline
if (!is.null(glb_Baseline_mdl_var))
ret_lst <- myfit_mdl(mdl_id="Baseline",
model_method="mybaseln_classfr",
indep_vars_vctr=glb_Baseline_mdl_var,
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glbObsFit, OOB_df=glbObsOOB)
# Most Frequent Outcome "MFO" model: mean(y) for regression
# Not using caret's nullModel since model stats not avl
# Cannot use rpart for multinomial classification since it predicts non-MFO
ret_lst <- myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = "MFO", type = glb_model_type, trainControl.method = "none",
train.method = ifelse(glb_is_regression, "lm", "myMFO_classfr"))),
indep_vars = ".rnorm", rsp_var = glb_rsp_var,
fit_df = glbObsFit, OOB_df = glbObsOOB)## [1] "fitting model: MFO.myMFO_classfr"
## [1] " indep_vars: .rnorm"
## Fitting parameter = none on full training set
## [1] "in MFO.Classifier$fit"
## [1] "unique.vals:"
## [1] N Y
## Levels: N Y
## [1] "unique.prob:"
## y
## N Y
## 0.820358 0.179642
## [1] "MFO.val:"
## [1] "N"
## Length Class Mode
## unique.vals 2 factor numeric
## unique.prob 2 -none- numeric
## MFO.val 1 -none- character
## x.names 1 -none- character
## xNames 1 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] "entr MFO.Classifier$predict"
## [1] "exit MFO.Classifier$predict"## Loading required package: ROCR
## Loading required package: gplots
##
## Attaching package: 'gplots'
##
## The following object is masked from 'package:stats':
##
## lowess## [1] "in MFO.Classifier$prob"
## N Y
## 1 0.820358 0.179642
## 2 0.820358 0.179642
## 3 0.820358 0.179642
## 4 0.820358 0.179642
## 5 0.820358 0.179642
## 6 0.820358 0.179642
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.Y
## 1 N 3941
## 2 Y 863
## Prediction
## Reference N Y
## N 0 3941
## Y 0 863
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1796420 0.0000000 0.1688795 0.1907952 0.8203580
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## [1] "entr MFO.Classifier$predict"
## [1] "exit MFO.Classifier$predict"
## [1] "in MFO.Classifier$prob"
## N Y
## 1 0.820358 0.179642
## 2 0.820358 0.179642
## 3 0.820358 0.179642
## 4 0.820358 0.179642
## 5 0.820358 0.179642
## 6 0.820358 0.179642
## Popular.fctr Popular.fctr.predict.MFO.myMFO_classfr.Y
## 1 N 1498
## 2 Y 230
## Prediction
## Reference N Y
## N 0 1498
## Y 0 230
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1331019 0.0000000 0.1174298 0.1500310 0.8668981
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## id feats max.nTuningRuns min.elapsedtime.everything
## 1 MFO.myMFO_classfr .rnorm 0 0.269
## min.elapsedtime.final max.AUCpROC.fit max.Sens.fit max.Spec.fit
## 1 0.003 0.5 1 0
## max.AUCROCR.fit opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0.1 0.3045703 0.179642
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.1688795 0.1907952 0
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5 1 0 0.5
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.2349336 0.1331019
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.1174298 0.150031 0if (glb_is_classification)
# "random" model - only for classification;
# none needed for regression since it is same as MFO
ret_lst <- myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = "Random", type = glb_model_type, trainControl.method = "none",
train.method = "myrandom_classfr")),
indep_vars = ".rnorm", rsp_var = glb_rsp_var,
fit_df = glbObsFit, OOB_df = glbObsOOB)## [1] "fitting model: Random.myrandom_classfr"
## [1] " indep_vars: .rnorm"
## Fitting parameter = none on full training set
## Length Class Mode
## unique.vals 2 factor numeric
## unique.prob 2 table numeric
## xNames 1 -none- character
## problemType 1 -none- character
## tuneValue 1 data.frame list
## obsLevels 2 -none- character
## [1] "in Random.Classifier$prob"
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 N 3941
## 2 Y 863
## Prediction
## Reference N Y
## N 0 3941
## Y 0 863
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1796420 0.0000000 0.1688795 0.1907952 0.8203580
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## [1] "in Random.Classifier$prob"
## Popular.fctr Popular.fctr.predict.Random.myrandom_classfr.Y
## 1 N 1498
## 2 Y 230
## Prediction
## Reference N Y
## N 0 1498
## Y 0 230
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.1331019 0.0000000 0.1174298 0.1500310 0.8668981
## AccuracyPValue McnemarPValue
## 1.0000000 0.0000000
## id feats max.nTuningRuns
## 1 Random.myrandom_classfr .rnorm 0
## min.elapsedtime.everything min.elapsedtime.final max.AUCpROC.fit
## 1 0.269 0.001 0.4990604
## max.Sens.fit max.Spec.fit max.AUCROCR.fit opt.prob.threshold.fit
## 1 0.8312611 0.1668598 0.4972757 0.1
## max.f.score.fit max.Accuracy.fit max.AccuracyLower.fit
## 1 0.3045703 0.179642 0.1688795
## max.AccuracyUpper.fit max.Kappa.fit max.AUCpROC.OOB max.Sens.OOB
## 1 0.1907952 0 0.5125675 0.8077437
## max.Spec.OOB max.AUCROCR.OOB opt.prob.threshold.OOB max.f.score.OOB
## 1 0.2173913 0.4857956 0.1 0.2349336
## max.Accuracy.OOB max.AccuracyLower.OOB max.AccuracyUpper.OOB
## 1 0.1331019 0.1174298 0.150031
## max.Kappa.OOB
## 1 0# ret_lst <- myfit_mdl(mdl_id = "Random", model_method = "myrandom_classfr",
# model_type = glb_model_type,
# indep_vars_vctr = ".rnorm",
# rsp_var = glb_rsp_var, rsp_var_out = glb_rsp_var_out,
# fit_df = glbObsFit, OOB_df = glbObsOOB)
# Max.cor.Y
# Check impact of cv
# rpart is not a good candidate since caret does not optimize cp (only tuning parameter of rpart) well
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix="Max.cor.Y.rcv.1X1", type=glb_model_type, trainControl.method="none",
train.method="glmnet")),
indep_vars=max_cor_y_x_vars, rsp_var=glb_rsp_var,
fit_df=glbObsFit, OOB_df=glbObsOOB)## [1] "fitting model: Max.cor.Y.rcv.1X1.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"## Loading required package: glmnet
## Loading required package: Matrix
## Loaded glmnet 2.0-2## Fitting alpha = 0.1, lambda = 0.00434 on full training set
## Length Class Mode
## a0 100 -none- numeric
## beta 2100 dgCMatrix S4
## df 100 -none- numeric
## dim 2 -none- numeric
## lambda 100 -none- numeric
## dev.ratio 100 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -4.57159198
## NDSSName.my.fctr#Multimedia#
## -1.22219085
## NDSSName.my.fctr#Opinion#RoomForDebate
## -3.46072453
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 4.06871185
## NDSSName.my.fctr#U.S.#Education
## -1.89443632
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22472818
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.95537118
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.55408513
## NDSSName.my.fctrBusiness#Technology#
## 0.77368538
## NDSSName.my.fctrCulture#Arts#
## -0.09465691
## NDSSName.my.fctrForeign#World#
## -1.45528874
## NDSSName.my.fctrForeign#World#AsiaPacific
## -1.60117505
## NDSSName.my.fctrMetro#N.Y./Region#
## 0.01563989
## NDSSName.my.fctrOpEd#Opinion#
## 4.51696382
## NDSSName.my.fctrScience#Health#
## 3.51595317
## NDSSName.my.fctrStyles##Fashion
## -1.85948925
## NDSSName.my.fctrStyles#U.S.#
## 3.27995325
## NDSSName.my.fctrTStyle##
## -1.54110404
## NDSSName.my.fctrTravel#Travel#
## -1.41940605
## NDSSName.my.fctrmyOther
## -1.90156922
## WordCount.root2
## 0.08434378
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -4.60394059
## NDSSName.my.fctr#Multimedia#
## -1.25163328
## NDSSName.my.fctr#Opinion#RoomForDebate
## -3.55521332
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 4.09217313
## NDSSName.my.fctr#U.S.#Education
## -1.96172971
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22495986
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.96836050
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.58120497
## NDSSName.my.fctrBusiness#Technology#
## 0.78504703
## NDSSName.my.fctrCulture#Arts#
## -0.09069661
## NDSSName.my.fctrForeign#World#
## -1.51061232
## NDSSName.my.fctrForeign#World#AsiaPacific
## -1.63313235
## NDSSName.my.fctrMetro#N.Y./Region#
## 0.02466697
## NDSSName.my.fctrOpEd#Opinion#
## 4.54361134
## NDSSName.my.fctrScience#Health#
## 3.53210055
## NDSSName.my.fctrStyles##Fashion
## -1.92188290
## NDSSName.my.fctrStyles#U.S.#
## 3.29488750
## NDSSName.my.fctrTStyle##
## -1.57788931
## NDSSName.my.fctrTravel#Travel#
## -1.47368131
## NDSSName.my.fctrmyOther
## -1.97357582
## WordCount.root2
## 0.08537319
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.1X1.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Max.cor.Y.rcv.1X1.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.1X1.glmnet.N
## 1 N 1151
## 2 Y 67
## Popular.fctr.predict.Max.cor.Y.rcv.1X1.glmnet.Y
## 1 347
## 2 163
## Prediction
## Reference N Y
## N 1151 347
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.604167e-01 3.148374e-01 7.395703e-01 7.803749e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 8.593187e-43
## id feats
## 1 Max.cor.Y.rcv.1X1.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 0 0.972 0.273
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8790544 0.9632073 0.7949015 0.9608594
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0.8099174 0.9329725
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7692476
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8116126
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4405405 0.7604167
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7395703 0.7803749 0.3148374# rcv_n_folds == 1 & rcv_n_repeats > 1 crashes
for (rcv_n_folds in seq(3, glb_rcv_n_folds + 2, 2))
for (rcv_n_repeats in seq(1, glb_rcv_n_repeats + 2, 2)) {
# Experiment specific code to avoid caret crash
# lcl_tune_models_df <- rbind(data.frame()
# ,data.frame(method = "glmnet", parameter = "alpha",
# vals = "0.100 0.325 0.550 0.775 1.000")
# ,data.frame(method = "glmnet", parameter = "lambda",
# vals = "9.342e-02")
# )
ret_lst <- myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst =
list(
id.prefix = paste0("Max.cor.Y.rcv.", rcv_n_folds, "X", rcv_n_repeats),
type = glb_model_type,
# tune.df = lcl_tune_models_df,
trainControl.method = "repeatedcv",
trainControl.number = rcv_n_folds,
trainControl.repeats = rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.method = "glmnet", train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize)),
indep_vars = max_cor_y_x_vars, rsp_var = glb_rsp_var,
fit_df = glbObsFit, OOB_df = glbObsOOB)
}## [1] "fitting model: Max.cor.Y.rcv.3X1.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0201 on full training set
## Length Class Mode
## a0 99 -none- numeric
## beta 2079 dgCMatrix S4
## df 99 -none- numeric
## dim 2 -none- numeric
## lambda 99 -none- numeric
## dev.ratio 99 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.89350373
## NDSSName.my.fctr#Multimedia#
## -0.01916344
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.18453357
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.21701058
## NDSSName.my.fctr#U.S.#Education
## -0.47679040
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.09891374
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.87281404
## NDSSName.my.fctrBusiness#Technology#
## 0.40965256
## NDSSName.my.fctrForeign#World#
## -0.05114617
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.47464340
## NDSSName.my.fctrOpEd#Opinion#
## 3.95214357
## NDSSName.my.fctrScience#Health#
## 3.14232408
## NDSSName.my.fctrStyles##Fashion
## -0.31867093
## NDSSName.my.fctrStyles#U.S.#
## 2.92610567
## NDSSName.my.fctrTStyle##
## -0.60538025
## WordCount.root2
## 0.05783392
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.95644632
## NDSSName.my.fctr#Multimedia#
## -0.07182859
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.30034382
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.29694236
## NDSSName.my.fctr#U.S.#Education
## -0.53415905
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.14259759
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.94231812
## NDSSName.my.fctrBusiness#Technology#
## 0.45657914
## NDSSName.my.fctrForeign#World#
## -0.10021084
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.53077048
## NDSSName.my.fctrOpEd#Opinion#
## 4.01324666
## NDSSName.my.fctrScience#Health#
## 3.18936803
## NDSSName.my.fctrStyles##Fashion
## -0.38069674
## NDSSName.my.fctrStyles#U.S.#
## 2.97176051
## NDSSName.my.fctrTStyle##
## -0.64837249
## WordCount.root2
## 0.05978318
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X1.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Max.cor.Y.rcv.3X1.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X1.glmnet.N
## 1 N 1146
## 2 Y 67
## Popular.fctr.predict.Max.cor.Y.rcv.3X1.glmnet.Y
## 1 352
## 2 163
## Prediction
## Reference N Y
## N 1146 352
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.575231e-01 3.107477e-01 7.365992e-01 7.775689e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 9.066396e-44
## id feats
## 1 Max.cor.Y.rcv.3X1.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 2.826 0.27
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8767919 0.964476 0.7891078 0.9582555
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8099174 0.9335973
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7691678
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8067975
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4375839 0.7575231
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7365992 0.7775689 0.3107477
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.007015493 0.02403706
## [1] "fitting model: Max.cor.Y.rcv.3X3.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0201 on full training set
## Length Class Mode
## a0 99 -none- numeric
## beta 2079 dgCMatrix S4
## df 99 -none- numeric
## dim 2 -none- numeric
## lambda 99 -none- numeric
## dev.ratio 99 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.89350373
## NDSSName.my.fctr#Multimedia#
## -0.01916344
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.18453357
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.21701058
## NDSSName.my.fctr#U.S.#Education
## -0.47679040
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.09891374
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.87281404
## NDSSName.my.fctrBusiness#Technology#
## 0.40965256
## NDSSName.my.fctrForeign#World#
## -0.05114617
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.47464340
## NDSSName.my.fctrOpEd#Opinion#
## 3.95214357
## NDSSName.my.fctrScience#Health#
## 3.14232408
## NDSSName.my.fctrStyles##Fashion
## -0.31867093
## NDSSName.my.fctrStyles#U.S.#
## 2.92610567
## NDSSName.my.fctrTStyle##
## -0.60538025
## WordCount.root2
## 0.05783392
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.95644632
## NDSSName.my.fctr#Multimedia#
## -0.07182859
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.30034382
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.29694236
## NDSSName.my.fctr#U.S.#Education
## -0.53415905
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.14259759
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.94231812
## NDSSName.my.fctrBusiness#Technology#
## 0.45657914
## NDSSName.my.fctrForeign#World#
## -0.10021084
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.53077048
## NDSSName.my.fctrOpEd#Opinion#
## 4.01324666
## NDSSName.my.fctrScience#Health#
## 3.18936803
## NDSSName.my.fctrStyles##Fashion
## -0.38069674
## NDSSName.my.fctrStyles#U.S.#
## 2.97176051
## NDSSName.my.fctrTStyle##
## -0.64837249
## WordCount.root2
## 0.05978318
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X3.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Max.cor.Y.rcv.3X3.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X3.glmnet.N
## 1 N 1146
## 2 Y 67
## Popular.fctr.predict.Max.cor.Y.rcv.3X3.glmnet.Y
## 1 352
## 2 163
## Prediction
## Reference N Y
## N 1146 352
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.575231e-01 3.107477e-01 7.365992e-01 7.775689e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 9.066396e-44
## id feats
## 1 Max.cor.Y.rcv.3X3.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 4.817 0.27
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8767919 0.964476 0.7891078 0.9582555
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8099174 0.9333193
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7690803
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8067975
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4375839 0.7575231
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7365992 0.7775689 0.3107477
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.005178375 0.01754365
## [1] "fitting model: Max.cor.Y.rcv.3X5.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0201 on full training set
## Length Class Mode
## a0 99 -none- numeric
## beta 2079 dgCMatrix S4
## df 99 -none- numeric
## dim 2 -none- numeric
## lambda 99 -none- numeric
## dev.ratio 99 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.89350373
## NDSSName.my.fctr#Multimedia#
## -0.01916344
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.18453357
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.21701058
## NDSSName.my.fctr#U.S.#Education
## -0.47679040
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.09891374
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.87281404
## NDSSName.my.fctrBusiness#Technology#
## 0.40965256
## NDSSName.my.fctrForeign#World#
## -0.05114617
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.47464340
## NDSSName.my.fctrOpEd#Opinion#
## 3.95214357
## NDSSName.my.fctrScience#Health#
## 3.14232408
## NDSSName.my.fctrStyles##Fashion
## -0.31867093
## NDSSName.my.fctrStyles#U.S.#
## 2.92610567
## NDSSName.my.fctrTStyle##
## -0.60538025
## WordCount.root2
## 0.05783392
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.95644632
## NDSSName.my.fctr#Multimedia#
## -0.07182859
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.30034382
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.29694236
## NDSSName.my.fctr#U.S.#Education
## -0.53415905
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.14259759
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.94231812
## NDSSName.my.fctrBusiness#Technology#
## 0.45657914
## NDSSName.my.fctrForeign#World#
## -0.10021084
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.53077048
## NDSSName.my.fctrOpEd#Opinion#
## 4.01324666
## NDSSName.my.fctrScience#Health#
## 3.18936803
## NDSSName.my.fctrStyles##Fashion
## -0.38069674
## NDSSName.my.fctrStyles#U.S.#
## 2.97176051
## NDSSName.my.fctrTStyle##
## -0.64837249
## WordCount.root2
## 0.05978318
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X5.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Max.cor.Y.rcv.3X5.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.3X5.glmnet.N
## 1 N 1146
## 2 Y 67
## Popular.fctr.predict.Max.cor.Y.rcv.3X5.glmnet.Y
## 1 352
## 2 163
## Prediction
## Reference N Y
## N 1146 352
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.575231e-01 3.107477e-01 7.365992e-01 7.775689e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 9.066396e-44
## id feats
## 1 Max.cor.Y.rcv.3X5.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 6.419 0.269
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8767919 0.964476 0.7891078 0.9582555
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8099174 0.9332218
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7686375
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8067975
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4375839 0.7575231
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7365992 0.7775689 0.3107477
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.005396525 0.01835474
## [1] "fitting model: Max.cor.Y.rcv.5X1.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.1, lambda = 0.0201 on full training set## Warning in myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst =
## list(id.prefix = paste0("Max.cor.Y.rcv.", : model's bestTune found at an
## extreme of tuneGrid for parameter: alpha
## Length Class Mode
## a0 100 -none- numeric
## beta 2100 dgCMatrix S4
## df 100 -none- numeric
## dim 2 -none- numeric
## lambda 100 -none- numeric
## dev.ratio 100 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.81141260
## NDSSName.my.fctr#Multimedia#
## -0.68105584
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.92624537
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.40699589
## NDSSName.my.fctr#U.S.#Education
## -0.98291999
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22577146
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.64343834
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.82797332
## NDSSName.my.fctrBusiness#Technology#
## 0.45317927
## NDSSName.my.fctrCulture#Arts#
## -0.17187706
## NDSSName.my.fctrForeign#World#
## -0.72035867
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.99018968
## NDSSName.my.fctrOpEd#Opinion#
## 3.81891156
## NDSSName.my.fctrScience#Health#
## 3.05516080
## NDSSName.my.fctrStyles##Fashion
## -0.97651721
## NDSSName.my.fctrStyles#U.S.#
## 2.84779285
## NDSSName.my.fctrTStyle##
## -0.94109645
## NDSSName.my.fctrTravel#Travel#
## -0.68827560
## NDSSName.my.fctrmyOther
## -0.84423735
## WordCount.root2
## 0.06115867
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.87108412
## NDSSName.my.fctr#Multimedia#
## -0.71588942
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.02010163
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.46715540
## NDSSName.my.fctr#U.S.#Education
## -1.02957582
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22558850
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.66798026
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.89132347
## NDSSName.my.fctrBusiness#Technology#
## 0.48212450
## NDSSName.my.fctrCulture#Arts#
## -0.16733777
## NDSSName.my.fctrForeign#World#
## -0.75793881
## NDSSName.my.fctrForeign#World#AsiaPacific
## -1.03076807
## NDSSName.my.fctrOpEd#Opinion#
## 3.87908175
## NDSSName.my.fctrScience#Health#
## 3.09788786
## NDSSName.my.fctrStyles##Fashion
## -1.02481879
## NDSSName.my.fctrStyles#U.S.#
## 2.88826078
## NDSSName.my.fctrTStyle##
## -0.97585470
## NDSSName.my.fctrTravel#Travel#
## -0.72668427
## NDSSName.my.fctrmyOther
## -0.90347045
## WordCount.root2
## 0.06289698
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X1.glmnet.N
## 1 N 3800
## 2 Y 179
## Popular.fctr.predict.Max.cor.Y.rcv.5X1.glmnet.Y
## 1 141
## 2 684
## Prediction
## Reference N Y
## N 3800 141
## Y 179 684
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.333888e-01 7.700473e-01 9.259666e-01 9.402789e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.097051e-115 3.860591e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X1.glmnet.N
## 1 N 1137
## 2 Y 53
## Popular.fctr.predict.Max.cor.Y.rcv.5X1.glmnet.Y
## 1 361
## 2 177
## Prediction
## Reference N Y
## N 1137 361
## Y 53 177
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.604167e-01 3.373693e-01 7.395703e-01 7.803749e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 1.935747e-51
## id feats
## 1 Max.cor.Y.rcv.5X1.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 3.223 0.267
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8784031 0.9642223 0.792584 0.9607052
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0.8104265 0.9331818
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9259666 0.9402789 0.7689055
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8114863
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4609375 0.7604167
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7395703 0.7803749 0.3373693
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.008837283 0.03133449
## [1] "fitting model: Max.cor.Y.rcv.5X3.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.1, lambda = 0.0201 on full training set## Warning in myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst =
## list(id.prefix = paste0("Max.cor.Y.rcv.", : model's bestTune found at an
## extreme of tuneGrid for parameter: alpha
## Length Class Mode
## a0 100 -none- numeric
## beta 2100 dgCMatrix S4
## df 100 -none- numeric
## dim 2 -none- numeric
## lambda 100 -none- numeric
## dev.ratio 100 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.81141260
## NDSSName.my.fctr#Multimedia#
## -0.68105584
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.92624537
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.40699589
## NDSSName.my.fctr#U.S.#Education
## -0.98291999
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22577146
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.64343834
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.82797332
## NDSSName.my.fctrBusiness#Technology#
## 0.45317927
## NDSSName.my.fctrCulture#Arts#
## -0.17187706
## NDSSName.my.fctrForeign#World#
## -0.72035867
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.99018968
## NDSSName.my.fctrOpEd#Opinion#
## 3.81891156
## NDSSName.my.fctrScience#Health#
## 3.05516080
## NDSSName.my.fctrStyles##Fashion
## -0.97651721
## NDSSName.my.fctrStyles#U.S.#
## 2.84779285
## NDSSName.my.fctrTStyle##
## -0.94109645
## NDSSName.my.fctrTravel#Travel#
## -0.68827560
## NDSSName.my.fctrmyOther
## -0.84423735
## WordCount.root2
## 0.06115867
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.87108412
## NDSSName.my.fctr#Multimedia#
## -0.71588942
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.02010163
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.46715540
## NDSSName.my.fctr#U.S.#Education
## -1.02957582
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22558850
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.66798026
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.89132347
## NDSSName.my.fctrBusiness#Technology#
## 0.48212450
## NDSSName.my.fctrCulture#Arts#
## -0.16733777
## NDSSName.my.fctrForeign#World#
## -0.75793881
## NDSSName.my.fctrForeign#World#AsiaPacific
## -1.03076807
## NDSSName.my.fctrOpEd#Opinion#
## 3.87908175
## NDSSName.my.fctrScience#Health#
## 3.09788786
## NDSSName.my.fctrStyles##Fashion
## -1.02481879
## NDSSName.my.fctrStyles#U.S.#
## 2.88826078
## NDSSName.my.fctrTStyle##
## -0.97585470
## NDSSName.my.fctrTravel#Travel#
## -0.72668427
## NDSSName.my.fctrmyOther
## -0.90347045
## WordCount.root2
## 0.06289698
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X3.glmnet.N
## 1 N 3800
## 2 Y 179
## Popular.fctr.predict.Max.cor.Y.rcv.5X3.glmnet.Y
## 1 141
## 2 684
## Prediction
## Reference N Y
## N 3800 141
## Y 179 684
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.333888e-01 7.700473e-01 9.259666e-01 9.402789e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.097051e-115 3.860591e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X3.glmnet.N
## 1 N 1137
## 2 Y 53
## Popular.fctr.predict.Max.cor.Y.rcv.5X3.glmnet.Y
## 1 361
## 2 177
## Prediction
## Reference N Y
## N 1137 361
## Y 53 177
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.604167e-01 3.373693e-01 7.395703e-01 7.803749e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 1.935747e-51
## id feats
## 1 Max.cor.Y.rcv.5X3.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 6.142 0.267
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8784031 0.9642223 0.792584 0.9607052
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0.8104265 0.9333905
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9259666 0.9402789 0.7698577
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8114863
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4609375 0.7604167
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7395703 0.7803749 0.3373693
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.006138477 0.02161286
## [1] "fitting model: Max.cor.Y.rcv.5X5.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.1, lambda = 0.0201 on full training set## Warning in myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst =
## list(id.prefix = paste0("Max.cor.Y.rcv.", : model's bestTune found at an
## extreme of tuneGrid for parameter: alpha
## Length Class Mode
## a0 100 -none- numeric
## beta 2100 dgCMatrix S4
## df 100 -none- numeric
## dim 2 -none- numeric
## lambda 100 -none- numeric
## dev.ratio 100 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.81141260
## NDSSName.my.fctr#Multimedia#
## -0.68105584
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.92624537
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.40699589
## NDSSName.my.fctr#U.S.#Education
## -0.98291999
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22577146
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.64343834
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.82797332
## NDSSName.my.fctrBusiness#Technology#
## 0.45317927
## NDSSName.my.fctrCulture#Arts#
## -0.17187706
## NDSSName.my.fctrForeign#World#
## -0.72035867
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.99018968
## NDSSName.my.fctrOpEd#Opinion#
## 3.81891156
## NDSSName.my.fctrScience#Health#
## 3.05516080
## NDSSName.my.fctrStyles##Fashion
## -0.97651721
## NDSSName.my.fctrStyles#U.S.#
## 2.84779285
## NDSSName.my.fctrTStyle##
## -0.94109645
## NDSSName.my.fctrTravel#Travel#
## -0.68827560
## NDSSName.my.fctrmyOther
## -0.84423735
## WordCount.root2
## 0.06115867
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.87108412
## NDSSName.my.fctr#Multimedia#
## -0.71588942
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.02010163
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.46715540
## NDSSName.my.fctr#U.S.#Education
## -1.02957582
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.22558850
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.66798026
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.89132347
## NDSSName.my.fctrBusiness#Technology#
## 0.48212450
## NDSSName.my.fctrCulture#Arts#
## -0.16733777
## NDSSName.my.fctrForeign#World#
## -0.75793881
## NDSSName.my.fctrForeign#World#AsiaPacific
## -1.03076807
## NDSSName.my.fctrOpEd#Opinion#
## 3.87908175
## NDSSName.my.fctrScience#Health#
## 3.09788786
## NDSSName.my.fctrStyles##Fashion
## -1.02481879
## NDSSName.my.fctrStyles#U.S.#
## 2.88826078
## NDSSName.my.fctrTStyle##
## -0.97585470
## NDSSName.my.fctrTravel#Travel#
## -0.72668427
## NDSSName.my.fctrmyOther
## -0.90347045
## WordCount.root2
## 0.06289698
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X5.glmnet.N
## 1 N 3800
## 2 Y 179
## Popular.fctr.predict.Max.cor.Y.rcv.5X5.glmnet.Y
## 1 141
## 2 684
## Prediction
## Reference N Y
## N 3800 141
## Y 179 684
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.333888e-01 7.700473e-01 9.259666e-01 9.402789e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.097051e-115 3.860591e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.5X5.glmnet.N
## 1 N 1137
## 2 Y 53
## Popular.fctr.predict.Max.cor.Y.rcv.5X5.glmnet.Y
## 1 361
## 2 177
## Prediction
## Reference N Y
## N 1137 361
## Y 53 177
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.604167e-01 3.373693e-01 7.395703e-01 7.803749e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 1.935747e-51
## id feats
## 1 Max.cor.Y.rcv.5X5.glmnet WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 8.908 0.268
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8784031 0.9642223 0.792584 0.9607052
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.5 0.8104265 0.9331816
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9259666 0.9402789 0.7691429
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8114863
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4609375 0.7604167
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7395703 0.7803749 0.3373693
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.0062138 0.02210061# Add parallel coordinates graph of glb_models_df[, glbMdlMetricsEval] to evaluate cv parameters
tmp_models_cols <- c("id", "max.nTuningRuns",
glbMdlMetricsEval[glbMdlMetricsEval %in% names(glb_models_df)],
grep("opt.", names(glb_models_df), fixed = TRUE, value = TRUE))
print(myplot_parcoord(obs_df = subset(glb_models_df,
grepl("Max.cor.Y.rcv.", id, fixed = TRUE),
select = -feats)[, tmp_models_cols],
id_var = "id"))
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix="Max.cor.Y.rcv.1X1.cp.0", type=glb_model_type, trainControl.method="none",
train.method="rpart",
tune.df=data.frame(method="rpart", parameter="cp", min=0.0, max=0.0, by=0.1))),
indep_vars=max_cor_y_x_vars, rsp_var=glb_rsp_var,
fit_df=glbObsFit, OOB_df=glbObsOOB)## [1] "fitting model: Max.cor.Y.rcv.1X1.cp.0.rpart"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"## Loading required package: rpart## Fitting cp = 0 on full training set## Loading required package: rpart.plot
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4804
##
## CP nsplit rel error
## 1 0.3696407879 0 1.0000000
## 2 0.0984936269 1 0.6303592
## 3 0.0857473928 2 0.5318656
## 4 0.0567786790 3 0.4461182
## 5 0.0104287370 4 0.3893395
## 6 0.0057937428 5 0.3789108
## 7 0.0034762457 7 0.3673233
## 8 0.0023174971 8 0.3638470
## 9 0.0011587486 11 0.3568946
## 10 0.0007724990 13 0.3545771
## 11 0.0005793743 16 0.3522596
## 12 0.0004213631 24 0.3476246
## 13 0.0003862495 35 0.3429896
## 14 0.0000000000 41 0.3406721
##
## Variable importance
## NDSSName.my.fctrOpEd#Opinion#
## 48
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 14
## NDSSName.my.fctrScience#Health#
## 14
## NDSSName.my.fctrStyles#U.S.#
## 11
## WordCount.root2
## 9
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 2
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## 1
##
## Node number 1: 4804 observations, complexity param=0.3696408
## predicted class=N expected loss=0.179642 P(node) =1
## class counts: 3941 863
## probabilities: 0.820 0.180
## left son=2 (4367 obs) right son=3 (437 obs)
## Primary splits:
## NDSSName.my.fctrOpEd#Opinion# < 0.5 to the left, improve=451.59770, (0 missing)
## NDSSName.my.fctrBusiness#Crosswords/Games# < 0.5 to the left, improve=112.88510, (0 missing)
## WordCount.root2 < 25.75849 to the left, improve=111.17610, (0 missing)
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve= 99.35206, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 68.73272, (0 missing)
##
## Node number 2: 4367 observations, complexity param=0.09849363
## predicted class=N expected loss=0.1110602 P(node) =0.9090341
## class counts: 3882 485
## probabilities: 0.889 0.111
## left son=4 (4262 obs) right son=5 (105 obs)
## Primary splits:
## NDSSName.my.fctrBusiness#Crosswords/Games# < 0.5 to the left, improve=135.55130, (0 missing)
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve=125.07920, (0 missing)
## WordCount.root2 < 25.75849 to the left, improve= 94.70710, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 88.56821, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 18.74400, (0 missing)
##
## Node number 3: 437 observations
## predicted class=Y expected loss=0.1350114 P(node) =0.09096586
## class counts: 59 378
## probabilities: 0.135 0.865
##
## Node number 4: 4262 observations, complexity param=0.08574739
## predicted class=N expected loss=0.09150634 P(node) =0.8871774
## class counts: 3872 390
## probabilities: 0.908 0.092
## left son=8 (4114 obs) right son=9 (148 obs)
## Primary splits:
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve=132.96710, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 94.69099, (0 missing)
## WordCount.root2 < 26.49528 to the left, improve= 84.07487, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 19.71762, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve= 10.17000, (0 missing)
##
## Node number 5: 105 observations, complexity param=0.002317497
## predicted class=Y expected loss=0.0952381 P(node) =0.02185679
## class counts: 10 95
## probabilities: 0.095 0.905
## left son=10 (12 obs) right son=11 (93 obs)
## Primary splits:
## WordCount.root2 < 18.9043 to the left, improve=6.455453, (0 missing)
##
## Node number 8: 4114 observations, complexity param=0.05677868
## predicted class=N expected loss=0.06781721 P(node) =0.8563697
## class counts: 3835 279
## probabilities: 0.932 0.068
## left son=16 (3987 obs) right son=17 (127 obs)
## Primary splits:
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve=102.410700, (0 missing)
## WordCount.root2 < 25.01 to the left, improve= 47.352210, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 20.930810, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve= 5.249425, (0 missing)
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve= 2.395935, (0 missing)
##
## Node number 9: 148 observations, complexity param=0.01042874
## predicted class=Y expected loss=0.25 P(node) =0.03080766
## class counts: 37 111
## probabilities: 0.250 0.750
## left son=18 (55 obs) right son=19 (93 obs)
## Primary splits:
## WordCount.root2 < 22.72663 to the left, improve=19.274, (0 missing)
##
## Node number 10: 12 observations
## predicted class=N expected loss=0.4166667 P(node) =0.002497918
## class counts: 7 5
## probabilities: 0.583 0.417
##
## Node number 11: 93 observations
## predicted class=Y expected loss=0.03225806 P(node) =0.01935887
## class counts: 3 90
## probabilities: 0.032 0.968
##
## Node number 16: 3987 observations, complexity param=0.005793743
## predicted class=N expected loss=0.04790569 P(node) =0.8299334
## class counts: 3796 191
## probabilities: 0.952 0.048
## left son=32 (2982 obs) right son=33 (1005 obs)
## Primary splits:
## WordCount.root2 < 25.01 to the left, improve=29.253580, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve=21.978920, (0 missing)
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve= 3.887348, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve= 2.348653, (0 missing)
## NDSSName.my.fctr#U.S.#Education < 0.5 to the right, improve= 1.187739, (0 missing)
## Surrogate splits:
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the left, agree=0.758, adj=0.042, (0 split)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the left, agree=0.752, adj=0.016, (0 split)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, agree=0.750, adj=0.008, (0 split)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the left, agree=0.748, adj=0.002, (0 split)
##
## Node number 17: 127 observations, complexity param=0.003476246
## predicted class=Y expected loss=0.3070866 P(node) =0.0264363
## class counts: 39 88
## probabilities: 0.307 0.693
## left son=34 (13 obs) right son=35 (114 obs)
## Primary splits:
## WordCount.root2 < 15.32846 to the left, improve=2.753047, (0 missing)
##
## Node number 18: 55 observations, complexity param=0.002317497
## predicted class=N expected loss=0.4181818 P(node) =0.01144879
## class counts: 32 23
## probabilities: 0.582 0.418
## left son=36 (9 obs) right son=37 (46 obs)
## Primary splits:
## WordCount.root2 < 19.93708 to the right, improve=0.8264383, (0 missing)
##
## Node number 19: 93 observations
## predicted class=Y expected loss=0.05376344 P(node) =0.01935887
## class counts: 5 88
## probabilities: 0.054 0.946
##
## Node number 32: 2982 observations
## predicted class=N expected loss=0.01274313 P(node) =0.6207327
## class counts: 2944 38
## probabilities: 0.987 0.013
##
## Node number 33: 1005 observations, complexity param=0.005793743
## predicted class=N expected loss=0.1522388 P(node) =0.2092007
## class counts: 852 153
## probabilities: 0.848 0.152
## left son=66 (993 obs) right son=67 (12 obs)
## Primary splits:
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve=14.193880, (0 missing)
## NDSSName.my.fctrCulture#Arts# < 0.5 to the left, improve= 3.669601, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve= 3.556158, (0 missing)
## WordCount.root2 < 34.19795 to the left, improve= 2.582851, (0 missing)
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the right, improve= 2.031748, (0 missing)
##
## Node number 34: 13 observations
## predicted class=N expected loss=0.3846154 P(node) =0.002706078
## class counts: 8 5
## probabilities: 0.615 0.385
##
## Node number 35: 114 observations, complexity param=0.000772499
## predicted class=Y expected loss=0.2719298 P(node) =0.02373022
## class counts: 31 83
## probabilities: 0.272 0.728
## left son=70 (79 obs) right son=71 (35 obs)
## Primary splits:
## WordCount.root2 < 29.21444 to the left, improve=1.020279, (0 missing)
##
## Node number 36: 9 observations
## predicted class=N expected loss=0.2222222 P(node) =0.001873439
## class counts: 7 2
## probabilities: 0.778 0.222
##
## Node number 37: 46 observations, complexity param=0.002317497
## predicted class=N expected loss=0.4565217 P(node) =0.009575354
## class counts: 25 21
## probabilities: 0.543 0.457
## left son=74 (36 obs) right son=75 (10 obs)
## Primary splits:
## WordCount.root2 < 17.01454 to the left, improve=1.514976, (0 missing)
##
## Node number 66: 993 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.143001 P(node) =0.2067027
## class counts: 851 142
## probabilities: 0.857 0.143
## left son=132 (930 obs) right son=133 (63 obs)
## Primary splits:
## NDSSName.my.fctrCulture#Arts# < 0.5 to the left, improve=4.094729, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=3.106316, (0 missing)
## WordCount.root2 < 29.5127 to the left, improve=2.722793, (0 missing)
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve=1.962300, (0 missing)
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the right, improve=1.793603, (0 missing)
##
## Node number 67: 12 observations
## predicted class=Y expected loss=0.08333333 P(node) =0.002497918
## class counts: 1 11
## probabilities: 0.083 0.917
##
## Node number 70: 79 observations, complexity param=0.000772499
## predicted class=Y expected loss=0.3164557 P(node) =0.01644463
## class counts: 25 54
## probabilities: 0.316 0.684
## left son=140 (25 obs) right son=141 (54 obs)
## Primary splits:
## WordCount.root2 < 27.36786 to the right, improve=0.5105485, (0 missing)
##
## Node number 71: 35 observations
## predicted class=Y expected loss=0.1714286 P(node) =0.007285595
## class counts: 6 29
## probabilities: 0.171 0.829
##
## Node number 74: 36 observations, complexity param=0.001158749
## predicted class=N expected loss=0.3888889 P(node) =0.007493755
## class counts: 22 14
## probabilities: 0.611 0.389
## left son=148 (8 obs) right son=149 (28 obs)
## Primary splits:
## WordCount.root2 < 15.74773 to the right, improve=0.3968254, (0 missing)
##
## Node number 75: 10 observations
## predicted class=Y expected loss=0.3 P(node) =0.002081599
## class counts: 3 7
## probabilities: 0.300 0.700
##
## Node number 132: 930 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.1311828 P(node) =0.1935887
## class counts: 808 122
## probabilities: 0.869 0.131
## left son=264 (627 obs) right son=265 (303 obs)
## Primary splits:
## WordCount.root2 < 33.97057 to the left, improve=2.913816, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=2.586923, (0 missing)
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve=2.402029, (0 missing)
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the right, improve=1.513920, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the left, improve=1.276783, (0 missing)
## Surrogate splits:
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the left, agree=0.719, adj=0.139, (0 split)
##
## Node number 133: 63 observations, complexity param=0.0003862495
## predicted class=N expected loss=0.3174603 P(node) =0.01311407
## class counts: 43 20
## probabilities: 0.683 0.317
## left son=266 (14 obs) right son=267 (49 obs)
## Primary splits:
## WordCount.root2 < 26.99984 to the left, improve=0.38322, (0 missing)
##
## Node number 140: 25 observations, complexity param=0.000772499
## predicted class=Y expected loss=0.4 P(node) =0.005203997
## class counts: 10 15
## probabilities: 0.400 0.600
## left son=280 (8 obs) right son=281 (17 obs)
## Primary splits:
## WordCount.root2 < 28.02674 to the left, improve=1.191176, (0 missing)
##
## Node number 141: 54 observations, complexity param=0.0003862495
## predicted class=Y expected loss=0.2777778 P(node) =0.01124063
## class counts: 15 39
## probabilities: 0.278 0.722
## left son=282 (45 obs) right son=283 (9 obs)
## Primary splits:
## WordCount.root2 < 26.55173 to the left, improve=0.6, (0 missing)
##
## Node number 148: 8 observations
## predicted class=N expected loss=0.25 P(node) =0.001665279
## class counts: 6 2
## probabilities: 0.750 0.250
##
## Node number 149: 28 observations, complexity param=0.001158749
## predicted class=N expected loss=0.4285714 P(node) =0.005828476
## class counts: 16 12
## probabilities: 0.571 0.429
## left son=298 (20 obs) right son=299 (8 obs)
## Primary splits:
## WordCount.root2 < 15.06648 to the left, improve=0.8642857, (0 missing)
##
## Node number 264: 627 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.1036683 P(node) =0.1305162
## class counts: 562 65
## probabilities: 0.896 0.104
## left son=528 (561 obs) right son=529 (66 obs)
## Primary splits:
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve=2.8404170, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=1.0796950, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve=1.0670160, (0 missing)
## WordCount.root2 < 29.5127 to the left, improve=0.8966879, (0 missing)
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, improve=0.4399337, (0 missing)
##
## Node number 265: 303 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.1881188 P(node) =0.06307244
## class counts: 246 57
## probabilities: 0.812 0.188
## left son=530 (222 obs) right son=531 (81 obs)
## Primary splits:
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the left, improve=5.4890570, (0 missing)
## WordCount.root2 < 38.17067 to the right, improve=5.0156320, (0 missing)
## NDSSName.my.fctr#Opinion#RoomForDebate < 0.5 to the right, improve=3.4510070, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=1.5155860, (0 missing)
## NDSSName.my.fctr#U.S.#Education < 0.5 to the right, improve=0.8078801, (0 missing)
## Surrogate splits:
## WordCount.root2 < 34.08078 to the right, agree=0.739, adj=0.025, (0 split)
##
## Node number 266: 14 observations
## predicted class=N expected loss=0.2142857 P(node) =0.002914238
## class counts: 11 3
## probabilities: 0.786 0.214
##
## Node number 267: 49 observations, complexity param=0.0003862495
## predicted class=N expected loss=0.3469388 P(node) =0.01019983
## class counts: 32 17
## probabilities: 0.653 0.347
## left son=534 (10 obs) right son=535 (39 obs)
## Primary splits:
## WordCount.root2 < 41.56249 to the right, improve=0.5425432, (0 missing)
##
## Node number 280: 8 observations
## predicted class=N expected loss=0.375 P(node) =0.001665279
## class counts: 5 3
## probabilities: 0.625 0.375
##
## Node number 281: 17 observations
## predicted class=Y expected loss=0.2941176 P(node) =0.003538718
## class counts: 5 12
## probabilities: 0.294 0.706
##
## Node number 282: 45 observations, complexity param=0.0003862495
## predicted class=Y expected loss=0.3111111 P(node) =0.009367194
## class counts: 14 31
## probabilities: 0.311 0.689
## left son=564 (23 obs) right son=565 (22 obs)
## Primary splits:
## WordCount.root2 < 21.70252 to the right, improve=0.6050944, (0 missing)
##
## Node number 283: 9 observations
## predicted class=Y expected loss=0.1111111 P(node) =0.001873439
## class counts: 1 8
## probabilities: 0.111 0.889
##
## Node number 298: 20 observations
## predicted class=N expected loss=0.35 P(node) =0.004163197
## class counts: 13 7
## probabilities: 0.650 0.350
##
## Node number 299: 8 observations
## predicted class=Y expected loss=0.375 P(node) =0.001665279
## class counts: 3 5
## probabilities: 0.375 0.625
##
## Node number 528: 561 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.08734403 P(node) =0.1167777
## class counts: 512 49
## probabilities: 0.913 0.087
## left son=1056 (281 obs) right son=1057 (280 obs)
## Primary splits:
## WordCount.root2 < 29.33428 to the left, improve=1.5853030, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve=0.7645570, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.7250433, (0 missing)
## NDSSName.my.fctrStyles##Fashion < 0.5 to the right, improve=0.3000638, (0 missing)
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, improve=0.2729836, (0 missing)
## Surrogate splits:
## NDSSName.my.fctrMetro#N.Y./Region# < 0.5 to the left, agree=0.560, adj=0.118, (0 split)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, agree=0.533, adj=0.064, (0 split)
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, agree=0.524, adj=0.046, (0 split)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the left, agree=0.515, adj=0.029, (0 split)
## NDSSName.my.fctrStyles##Fashion < 0.5 to the right, agree=0.512, adj=0.021, (0 split)
##
## Node number 529: 66 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.2424242 P(node) =0.01373855
## class counts: 50 16
## probabilities: 0.758 0.242
## left son=1058 (38 obs) right son=1059 (28 obs)
## Primary splits:
## WordCount.root2 < 27.86575 to the left, improve=0.6070859, (0 missing)
##
## Node number 530: 222 observations
## predicted class=N expected loss=0.1306306 P(node) =0.04621149
## class counts: 193 29
## probabilities: 0.869 0.131
##
## Node number 531: 81 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.345679 P(node) =0.01686095
## class counts: 53 28
## probabilities: 0.654 0.346
## left son=1062 (15 obs) right son=1063 (66 obs)
## Primary splits:
## WordCount.root2 < 41.59766 to the right, improve=2.866218, (0 missing)
##
## Node number 534: 10 observations
## predicted class=N expected loss=0.2 P(node) =0.002081599
## class counts: 8 2
## probabilities: 0.800 0.200
##
## Node number 535: 39 observations, complexity param=0.0003862495
## predicted class=N expected loss=0.3846154 P(node) =0.008118235
## class counts: 24 15
## probabilities: 0.615 0.385
## left son=1070 (32 obs) right son=1071 (7 obs)
## Primary splits:
## WordCount.root2 < 34.23387 to the left, improve=0.595467, (0 missing)
##
## Node number 564: 23 observations, complexity param=0.0003862495
## predicted class=Y expected loss=0.3913043 P(node) =0.004787677
## class counts: 9 14
## probabilities: 0.391 0.609
## left son=1128 (7 obs) right son=1129 (16 obs)
## Primary splits:
## WordCount.root2 < 23.6326 to the left, improve=0.6529503, (0 missing)
##
## Node number 565: 22 observations
## predicted class=Y expected loss=0.2272727 P(node) =0.004579517
## class counts: 5 17
## probabilities: 0.227 0.773
##
## Node number 1056: 281 observations
## predicted class=N expected loss=0.04982206 P(node) =0.05849292
## class counts: 267 14
## probabilities: 0.950 0.050
##
## Node number 1057: 280 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.125 P(node) =0.05828476
## class counts: 245 35
## probabilities: 0.875 0.125
## left son=2114 (71 obs) right son=2115 (209 obs)
## Primary splits:
## WordCount.root2 < 32.57299 to the right, improve=0.8968765, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve=0.7830739, (0 missing)
## NDSSName.my.fctrMetro#N.Y./Region# < 0.5 to the right, improve=0.3683673, (0 missing)
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, improve=0.3578067, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.3021494, (0 missing)
##
## Node number 1058: 38 observations
## predicted class=N expected loss=0.1842105 P(node) =0.007910075
## class counts: 31 7
## probabilities: 0.816 0.184
##
## Node number 1059: 28 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.3214286 P(node) =0.005828476
## class counts: 19 9
## probabilities: 0.679 0.321
## left son=2118 (19 obs) right son=2119 (9 obs)
## Primary splits:
## WordCount.root2 < 28.6269 to the right, improve=1.454052, (0 missing)
##
## Node number 1062: 15 observations
## predicted class=N expected loss=0.06666667 P(node) =0.003122398
## class counts: 14 1
## probabilities: 0.933 0.067
##
## Node number 1063: 66 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.4090909 P(node) =0.01373855
## class counts: 39 27
## probabilities: 0.591 0.409
## left son=2126 (25 obs) right son=2127 (41 obs)
## Primary splits:
## WordCount.root2 < 35.6581 to the left, improve=1.341286, (0 missing)
##
## Node number 1070: 32 observations
## predicted class=N expected loss=0.34375 P(node) =0.006661116
## class counts: 21 11
## probabilities: 0.656 0.344
##
## Node number 1071: 7 observations
## predicted class=Y expected loss=0.4285714 P(node) =0.001457119
## class counts: 3 4
## probabilities: 0.429 0.571
##
## Node number 1128: 7 observations
## predicted class=N expected loss=0.4285714 P(node) =0.001457119
## class counts: 4 3
## probabilities: 0.571 0.429
##
## Node number 1129: 16 observations
## predicted class=Y expected loss=0.3125 P(node) =0.003330558
## class counts: 5 11
## probabilities: 0.312 0.688
##
## Node number 2114: 71 observations
## predicted class=N expected loss=0.05633803 P(node) =0.01477935
## class counts: 67 4
## probabilities: 0.944 0.056
##
## Node number 2115: 209 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.1483254 P(node) =0.04350541
## class counts: 178 31
## probabilities: 0.852 0.148
## left son=4230 (12 obs) right son=4231 (197 obs)
## Primary splits:
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve=0.5601729, (0 missing)
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, improve=0.5108985, (0 missing)
## WordCount.root2 < 30.09153 to the right, improve=0.4980706, (0 missing)
## NDSSName.my.fctrMetro#N.Y./Region# < 0.5 to the right, improve=0.4241343, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.3390226, (0 missing)
##
## Node number 2118: 19 observations
## predicted class=N expected loss=0.2105263 P(node) =0.003955037
## class counts: 15 4
## probabilities: 0.789 0.211
##
## Node number 2119: 9 observations
## predicted class=Y expected loss=0.4444444 P(node) =0.001873439
## class counts: 4 5
## probabilities: 0.444 0.556
##
## Node number 2126: 25 observations
## predicted class=N expected loss=0.28 P(node) =0.005203997
## class counts: 18 7
## probabilities: 0.720 0.280
##
## Node number 2127: 41 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.4878049 P(node) =0.008534555
## class counts: 21 20
## probabilities: 0.512 0.488
## left son=4254 (30 obs) right son=4255 (11 obs)
## Primary splits:
## WordCount.root2 < 36.31791 to the right, improve=0.6635625, (0 missing)
##
## Node number 4230: 12 observations
## predicted class=N expected loss=0 P(node) =0.002497918
## class counts: 12 0
## probabilities: 1.000 0.000
##
## Node number 4231: 197 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.1573604 P(node) =0.04100749
## class counts: 166 31
## probabilities: 0.843 0.157
## left son=8462 (11 obs) right son=8463 (186 obs)
## Primary splits:
## NDSSName.my.fctr#Multimedia# < 0.5 to the right, improve=0.5769882, (0 missing)
## NDSSName.my.fctrMetro#N.Y./Region# < 0.5 to the right, improve=0.5314217, (0 missing)
## WordCount.root2 < 30.09153 to the right, improve=0.4682049, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.4106319, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the right, improve=0.1814254, (0 missing)
##
## Node number 4254: 30 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.4333333 P(node) =0.006244796
## class counts: 17 13
## probabilities: 0.567 0.433
## left son=8508 (7 obs) right son=8509 (23 obs)
## Primary splits:
## WordCount.root2 < 37.14159 to the left, improve=0.3979296, (0 missing)
##
## Node number 4255: 11 observations
## predicted class=Y expected loss=0.3636364 P(node) =0.002289759
## class counts: 4 7
## probabilities: 0.364 0.636
##
## Node number 8462: 11 observations
## predicted class=N expected loss=0 P(node) =0.002289759
## class counts: 11 0
## probabilities: 1.000 0.000
##
## Node number 8463: 186 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.1666667 P(node) =0.03871774
## class counts: 155 31
## probabilities: 0.833 0.167
## left son=16926 (29 obs) right son=16927 (157 obs)
## Primary splits:
## NDSSName.my.fctrMetro#N.Y./Region# < 0.5 to the right, improve=0.6559045, (0 missing)
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.4920635, (0 missing)
## WordCount.root2 < 30.09153 to the right, improve=0.3890196, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the right, improve=0.2415584, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the left, improve=0.0126479, (0 missing)
##
## Node number 8508: 7 observations
## predicted class=N expected loss=0.2857143 P(node) =0.001457119
## class counts: 5 2
## probabilities: 0.714 0.286
##
## Node number 8509: 23 observations, complexity param=0.0005793743
## predicted class=N expected loss=0.4782609 P(node) =0.004787677
## class counts: 12 11
## probabilities: 0.522 0.478
## left son=17018 (8 obs) right son=17019 (15 obs)
## Primary splits:
## WordCount.root2 < 38.57459 to the right, improve=0.2615942, (0 missing)
##
## Node number 16926: 29 observations
## predicted class=N expected loss=0.06896552 P(node) =0.006036636
## class counts: 27 2
## probabilities: 0.931 0.069
##
## Node number 16927: 157 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.1847134 P(node) =0.0326811
## class counts: 128 29
## probabilities: 0.815 0.185
## left son=33854 (18 obs) right son=33855 (139 obs)
## Primary splits:
## NDSSName.my.fctrForeign#World#AsiaPacific < 0.5 to the right, improve=0.67831090, (0 missing)
## WordCount.root2 < 32.38827 to the left, improve=0.61044970, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the right, improve=0.38816480, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the right, improve=0.01539613, (0 missing)
## Surrogate splits:
## WordCount.root2 < 32.51922 to the right, agree=0.892, adj=0.056, (0 split)
##
## Node number 17018: 8 observations
## predicted class=N expected loss=0.375 P(node) =0.001665279
## class counts: 5 3
## probabilities: 0.625 0.375
##
## Node number 17019: 15 observations
## predicted class=Y expected loss=0.4666667 P(node) =0.003122398
## class counts: 7 8
## probabilities: 0.467 0.533
##
## Node number 33854: 18 observations
## predicted class=N expected loss=0.05555556 P(node) =0.003746878
## class counts: 17 1
## probabilities: 0.944 0.056
##
## Node number 33855: 139 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.2014388 P(node) =0.02893422
## class counts: 111 28
## probabilities: 0.799 0.201
## left son=67710 (102 obs) right son=67711 (37 obs)
## Primary splits:
## WordCount.root2 < 30.09153 to the right, improve=0.9266317, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness < 0.5 to the right, improve=0.5580040, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the right, improve=0.1306354, (0 missing)
##
## Node number 67710: 102 observations
## predicted class=N expected loss=0.1666667 P(node) =0.02123231
## class counts: 85 17
## probabilities: 0.833 0.167
##
## Node number 67711: 37 observations, complexity param=0.0004213631
## predicted class=N expected loss=0.2972973 P(node) =0.007701915
## class counts: 26 11
## probabilities: 0.703 0.297
## left son=135422 (30 obs) right son=135423 (7 obs)
## Primary splits:
## WordCount.root2 < 29.92488 to the left, improve=3.00231700, (0 missing)
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook < 0.5 to the left, improve=0.01303089, (0 missing)
##
## Node number 135422: 30 observations
## predicted class=N expected loss=0.2 P(node) =0.006244796
## class counts: 24 6
## probabilities: 0.800 0.200
##
## Node number 135423: 7 observations
## predicted class=Y expected loss=0.2857143 P(node) =0.001457119
## class counts: 2 5
## probabilities: 0.286 0.714
##
## n= 4804
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4804 863 N (0.82035803 0.17964197)
## 2) NDSSName.my.fctrOpEd#Opinion#< 0.5 4367 485 N (0.88893978 0.11106022)
## 4) NDSSName.my.fctrBusiness#Crosswords/Games#< 0.5 4262 390 N (0.90849366 0.09150634)
## 8) NDSSName.my.fctrScience#Health#< 0.5 4114 279 N (0.93218279 0.06781721)
## 16) NDSSName.my.fctrStyles#U.S.#< 0.5 3987 191 N (0.95209431 0.04790569)
## 32) WordCount.root2< 25.01 2982 38 N (0.98725687 0.01274313) *
## 33) WordCount.root2>=25.01 1005 153 N (0.84776119 0.15223881)
## 66) NDSSName.my.fctr#Opinion#ThePublicEditor< 0.5 993 142 N (0.85699899 0.14300101)
## 132) NDSSName.my.fctrCulture#Arts#< 0.5 930 122 N (0.86881720 0.13118280)
## 264) WordCount.root2< 33.97057 627 65 N (0.89633174 0.10366826)
## 528) NDSSName.my.fctrBusiness#Technology#< 0.5 561 49 N (0.91265597 0.08734403)
## 1056) WordCount.root2< 29.33428 281 14 N (0.95017794 0.04982206) *
## 1057) WordCount.root2>=29.33428 280 35 N (0.87500000 0.12500000)
## 2114) WordCount.root2>=32.57299 71 4 N (0.94366197 0.05633803) *
## 2115) WordCount.root2< 32.57299 209 31 N (0.85167464 0.14832536)
## 4230) NDSSName.my.fctrTStyle##>=0.5 12 0 N (1.00000000 0.00000000) *
## 4231) NDSSName.my.fctrTStyle##< 0.5 197 31 N (0.84263959 0.15736041)
## 8462) NDSSName.my.fctr#Multimedia#>=0.5 11 0 N (1.00000000 0.00000000) *
## 8463) NDSSName.my.fctr#Multimedia#< 0.5 186 31 N (0.83333333 0.16666667)
## 16926) NDSSName.my.fctrMetro#N.Y./Region#>=0.5 29 2 N (0.93103448 0.06896552) *
## 16927) NDSSName.my.fctrMetro#N.Y./Region#< 0.5 157 29 N (0.81528662 0.18471338)
## 33854) NDSSName.my.fctrForeign#World#AsiaPacific>=0.5 18 1 N (0.94444444 0.05555556) *
## 33855) NDSSName.my.fctrForeign#World#AsiaPacific< 0.5 139 28 N (0.79856115 0.20143885)
## 67710) WordCount.root2>=30.09153 102 17 N (0.83333333 0.16666667) *
## 67711) WordCount.root2< 30.09153 37 11 N (0.70270270 0.29729730)
## 135422) WordCount.root2< 29.92488 30 6 N (0.80000000 0.20000000) *
## 135423) WordCount.root2>=29.92488 7 2 Y (0.28571429 0.71428571) *
## 529) NDSSName.my.fctrBusiness#Technology#>=0.5 66 16 N (0.75757576 0.24242424)
## 1058) WordCount.root2< 27.86575 38 7 N (0.81578947 0.18421053) *
## 1059) WordCount.root2>=27.86575 28 9 N (0.67857143 0.32142857)
## 2118) WordCount.root2>=28.6269 19 4 N (0.78947368 0.21052632) *
## 2119) WordCount.root2< 28.6269 9 4 Y (0.44444444 0.55555556) *
## 265) WordCount.root2>=33.97057 303 57 N (0.81188119 0.18811881)
## 530) NDSSName.my.fctrBusiness#BusinessDay#Dealbook< 0.5 222 29 N (0.86936937 0.13063063) *
## 531) NDSSName.my.fctrBusiness#BusinessDay#Dealbook>=0.5 81 28 N (0.65432099 0.34567901)
## 1062) WordCount.root2>=41.59766 15 1 N (0.93333333 0.06666667) *
## 1063) WordCount.root2< 41.59766 66 27 N (0.59090909 0.40909091)
## 2126) WordCount.root2< 35.6581 25 7 N (0.72000000 0.28000000) *
## 2127) WordCount.root2>=35.6581 41 20 N (0.51219512 0.48780488)
## 4254) WordCount.root2>=36.31791 30 13 N (0.56666667 0.43333333)
## 8508) WordCount.root2< 37.14159 7 2 N (0.71428571 0.28571429) *
## 8509) WordCount.root2>=37.14159 23 11 N (0.52173913 0.47826087)
## 17018) WordCount.root2>=38.57459 8 3 N (0.62500000 0.37500000) *
## 17019) WordCount.root2< 38.57459 15 7 Y (0.46666667 0.53333333) *
## 4255) WordCount.root2< 36.31791 11 4 Y (0.36363636 0.63636364) *
## 133) NDSSName.my.fctrCulture#Arts#>=0.5 63 20 N (0.68253968 0.31746032)
## 266) WordCount.root2< 26.99984 14 3 N (0.78571429 0.21428571) *
## 267) WordCount.root2>=26.99984 49 17 N (0.65306122 0.34693878)
## 534) WordCount.root2>=41.56249 10 2 N (0.80000000 0.20000000) *
## 535) WordCount.root2< 41.56249 39 15 N (0.61538462 0.38461538)
## 1070) WordCount.root2< 34.23387 32 11 N (0.65625000 0.34375000) *
## 1071) WordCount.root2>=34.23387 7 3 Y (0.42857143 0.57142857) *
## 67) NDSSName.my.fctr#Opinion#ThePublicEditor>=0.5 12 1 Y (0.08333333 0.91666667) *
## 17) NDSSName.my.fctrStyles#U.S.#>=0.5 127 39 Y (0.30708661 0.69291339)
## 34) WordCount.root2< 15.32846 13 5 N (0.61538462 0.38461538) *
## 35) WordCount.root2>=15.32846 114 31 Y (0.27192982 0.72807018)
## 70) WordCount.root2< 29.21444 79 25 Y (0.31645570 0.68354430)
## 140) WordCount.root2>=27.36786 25 10 Y (0.40000000 0.60000000)
## 280) WordCount.root2< 28.02674 8 3 N (0.62500000 0.37500000) *
## 281) WordCount.root2>=28.02674 17 5 Y (0.29411765 0.70588235) *
## 141) WordCount.root2< 27.36786 54 15 Y (0.27777778 0.72222222)
## 282) WordCount.root2< 26.55173 45 14 Y (0.31111111 0.68888889)
## 564) WordCount.root2>=21.70252 23 9 Y (0.39130435 0.60869565)
## 1128) WordCount.root2< 23.6326 7 3 N (0.57142857 0.42857143) *
## 1129) WordCount.root2>=23.6326 16 5 Y (0.31250000 0.68750000) *
## 565) WordCount.root2< 21.70252 22 5 Y (0.22727273 0.77272727) *
## 283) WordCount.root2>=26.55173 9 1 Y (0.11111111 0.88888889) *
## 71) WordCount.root2>=29.21444 35 6 Y (0.17142857 0.82857143) *
## 9) NDSSName.my.fctrScience#Health#>=0.5 148 37 Y (0.25000000 0.75000000)
## 18) WordCount.root2< 22.72663 55 23 N (0.58181818 0.41818182)
## 36) WordCount.root2>=19.93708 9 2 N (0.77777778 0.22222222) *
## 37) WordCount.root2< 19.93708 46 21 N (0.54347826 0.45652174)
## 74) WordCount.root2< 17.01454 36 14 N (0.61111111 0.38888889)
## 148) WordCount.root2>=15.74773 8 2 N (0.75000000 0.25000000) *
## 149) WordCount.root2< 15.74773 28 12 N (0.57142857 0.42857143)
## 298) WordCount.root2< 15.06648 20 7 N (0.65000000 0.35000000) *
## 299) WordCount.root2>=15.06648 8 3 Y (0.37500000 0.62500000) *
## 75) WordCount.root2>=17.01454 10 3 Y (0.30000000 0.70000000) *
## 19) WordCount.root2>=22.72663 93 5 Y (0.05376344 0.94623656) *
## 5) NDSSName.my.fctrBusiness#Crosswords/Games#>=0.5 105 10 Y (0.09523810 0.90476190)
## 10) WordCount.root2< 18.9043 12 5 N (0.58333333 0.41666667) *
## 11) WordCount.root2>=18.9043 93 3 Y (0.03225806 0.96774194) *
## 3) NDSSName.my.fctrOpEd#Opinion#>=0.5 437 59 Y (0.13501144 0.86498856) *
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.1X1.cp.0.rpart.N
## 1 N 3814
## 2 Y 170
## Popular.fctr.predict.Max.cor.Y.rcv.1X1.cp.0.rpart.Y
## 1 127
## 2 693
## Prediction
## Reference N Y
## N 3814 127
## Y 170 693
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.381765e-01 7.860827e-01 9.309917e-01 9.448229e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 2.798570e-127 1.480611e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rcv.1X1.cp.0.rpart.N
## 1 N 1180
## 2 Y 84
## Popular.fctr.predict.Max.cor.Y.rcv.1X1.cp.0.rpart.Y
## 1 318
## 2 146
## Prediction
## Reference N Y
## N 1180 318
## Y 84 146
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.673611e-01 2.953321e-01 7.467059e-01 7.871043e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 3.224022e-31
## id feats
## 1 Max.cor.Y.rcv.1X1.cp.0.rpart WordCount.root2,NDSSName.my.fctr
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 0 0.863 0.07
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8821543 0.9705658 0.7937428 0.9504198
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8235294 0.9381765
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9309917 0.9448229 0.7860827
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.6174697 0.9218959 0.3130435 0.7773858
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4207493 0.7673611
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7467059 0.7871043 0.2953321#stop(here"); glb_to_sav(); all.equal(glb_models_df, sav_models_df)
# if (glb_is_regression || glb_is_binomial) # For multinomials this model will be run next by default
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix="Max.cor.Y",
type=glb_model_type, trainControl.method="repeatedcv",
trainControl.number=glb_rcv_n_folds, trainControl.repeats=glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
trainControl.allowParallel = FALSE,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method="rpart")),
indep_vars=max_cor_y_x_vars, rsp_var=glb_rsp_var,
fit_df=glbObsFit, OOB_df=glbObsOOB)## [1] "fitting model: Max.cor.Y.rpart"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr"
## + Fold1.Rep1: cp=0.01043
## - Fold1.Rep1: cp=0.01043
## + Fold2.Rep1: cp=0.01043
## - Fold2.Rep1: cp=0.01043
## + Fold3.Rep1: cp=0.01043
## - Fold3.Rep1: cp=0.01043
## + Fold1.Rep2: cp=0.01043
## - Fold1.Rep2: cp=0.01043
## + Fold2.Rep2: cp=0.01043
## - Fold2.Rep2: cp=0.01043
## + Fold3.Rep2: cp=0.01043
## - Fold3.Rep2: cp=0.01043
## + Fold1.Rep3: cp=0.01043
## - Fold1.Rep3: cp=0.01043
## + Fold2.Rep3: cp=0.01043
## - Fold2.Rep3: cp=0.01043
## + Fold3.Rep3: cp=0.01043
## - Fold3.Rep3: cp=0.01043
## Aggregating results
## Selecting tuning parameters
## Fitting cp = 0.0104 on full training set## Warning in myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst
## = list(id.prefix = "Max.cor.Y", : model's bestTune found at an extreme of
## tuneGrid for parameter: cp
## Call:
## rpart(formula = .outcome ~ ., control = list(minsplit = 20, minbucket = 7,
## cp = 0, maxcompete = 4, maxsurrogate = 5, usesurrogate = 2,
## surrogatestyle = 0, maxdepth = 30, xval = 0))
## n= 4804
##
## CP nsplit rel error
## 1 0.36964079 0 1.0000000
## 2 0.09849363 1 0.6303592
## 3 0.08574739 2 0.5318656
## 4 0.05677868 3 0.4461182
## 5 0.01042874 4 0.3893395
##
## Variable importance
## NDSSName.my.fctrOpEd#Opinion#
## 55
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 16
## NDSSName.my.fctrScience#Health#
## 16
## NDSSName.my.fctrStyles#U.S.#
## 12
##
## Node number 1: 4804 observations, complexity param=0.3696408
## predicted class=N expected loss=0.179642 P(node) =1
## class counts: 3941 863
## probabilities: 0.820 0.180
## left son=2 (4367 obs) right son=3 (437 obs)
## Primary splits:
## NDSSName.my.fctrOpEd#Opinion# < 0.5 to the left, improve=451.59770, (0 missing)
## NDSSName.my.fctrBusiness#Crosswords/Games# < 0.5 to the left, improve=112.88510, (0 missing)
## WordCount.root2 < 25.75849 to the left, improve=111.17610, (0 missing)
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve= 99.35206, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 68.73272, (0 missing)
##
## Node number 2: 4367 observations, complexity param=0.09849363
## predicted class=N expected loss=0.1110602 P(node) =0.9090341
## class counts: 3882 485
## probabilities: 0.889 0.111
## left son=4 (4262 obs) right son=5 (105 obs)
## Primary splits:
## NDSSName.my.fctrBusiness#Crosswords/Games# < 0.5 to the left, improve=135.55130, (0 missing)
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve=125.07920, (0 missing)
## WordCount.root2 < 25.75849 to the left, improve= 94.70710, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 88.56821, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 18.74400, (0 missing)
##
## Node number 3: 437 observations
## predicted class=Y expected loss=0.1350114 P(node) =0.09096586
## class counts: 59 378
## probabilities: 0.135 0.865
##
## Node number 4: 4262 observations, complexity param=0.08574739
## predicted class=N expected loss=0.09150634 P(node) =0.8871774
## class counts: 3872 390
## probabilities: 0.908 0.092
## left son=8 (4114 obs) right son=9 (148 obs)
## Primary splits:
## NDSSName.my.fctrScience#Health# < 0.5 to the left, improve=132.96710, (0 missing)
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve= 94.69099, (0 missing)
## WordCount.root2 < 26.49528 to the left, improve= 84.07487, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 19.71762, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve= 10.17000, (0 missing)
##
## Node number 5: 105 observations
## predicted class=Y expected loss=0.0952381 P(node) =0.02185679
## class counts: 10 95
## probabilities: 0.095 0.905
##
## Node number 8: 4114 observations, complexity param=0.05677868
## predicted class=N expected loss=0.06781721 P(node) =0.8563697
## class counts: 3835 279
## probabilities: 0.932 0.068
## left son=16 (3987 obs) right son=17 (127 obs)
## Primary splits:
## NDSSName.my.fctrStyles#U.S.# < 0.5 to the left, improve=102.410700, (0 missing)
## WordCount.root2 < 25.01 to the left, improve= 47.352210, (0 missing)
## NDSSName.my.fctr#Opinion#ThePublicEditor < 0.5 to the left, improve= 20.930810, (0 missing)
## NDSSName.my.fctrTStyle## < 0.5 to the right, improve= 5.249425, (0 missing)
## NDSSName.my.fctrBusiness#Technology# < 0.5 to the left, improve= 2.395935, (0 missing)
##
## Node number 9: 148 observations
## predicted class=Y expected loss=0.25 P(node) =0.03080766
## class counts: 37 111
## probabilities: 0.250 0.750
##
## Node number 16: 3987 observations
## predicted class=N expected loss=0.04790569 P(node) =0.8299334
## class counts: 3796 191
## probabilities: 0.952 0.048
##
## Node number 17: 127 observations
## predicted class=Y expected loss=0.3070866 P(node) =0.0264363
## class counts: 39 88
## probabilities: 0.307 0.693
##
## n= 4804
##
## node), split, n, loss, yval, (yprob)
## * denotes terminal node
##
## 1) root 4804 863 N (0.82035803 0.17964197)
## 2) NDSSName.my.fctrOpEd#Opinion#< 0.5 4367 485 N (0.88893978 0.11106022)
## 4) NDSSName.my.fctrBusiness#Crosswords/Games#< 0.5 4262 390 N (0.90849366 0.09150634)
## 8) NDSSName.my.fctrScience#Health#< 0.5 4114 279 N (0.93218279 0.06781721)
## 16) NDSSName.my.fctrStyles#U.S.#< 0.5 3987 191 N (0.95209431 0.04790569) *
## 17) NDSSName.my.fctrStyles#U.S.#>=0.5 127 39 Y (0.30708661 0.69291339) *
## 9) NDSSName.my.fctrScience#Health#>=0.5 148 37 Y (0.25000000 0.75000000) *
## 5) NDSSName.my.fctrBusiness#Crosswords/Games#>=0.5 105 10 Y (0.09523810 0.90476190) *
## 3) NDSSName.my.fctrOpEd#Opinion#>=0.5 437 59 Y (0.13501144 0.86498856) *
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 3796
## 2 Y 191
## Popular.fctr.predict.Max.cor.Y.rpart.Y
## 1 145
## 2 672
## Prediction
## Reference N Y
## N 3796 145
## Y 191 672
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.300583e-01 7.576571e-01 9.224771e-01 9.371115e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 4.458834e-108 1.409037e-02
## Popular.fctr Popular.fctr.predict.Max.cor.Y.rpart.N
## 1 N 1355
## 2 Y 168
## Popular.fctr.predict.Max.cor.Y.rpart.Y
## 1 143
## 2 62
## Prediction
## Reference N Y
## N 1355 143
## Y 168 62
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 0.8200231 0.1825002 0.8010821 0.8378705 0.8668981
## AccuracyPValue McnemarPValue
## 1.0000000 0.1735405
## id feats max.nTuningRuns
## 1 Max.cor.Y.rpart WordCount.root2,NDSSName.my.fctr 5
## min.elapsedtime.everything min.elapsedtime.final max.AUCpROC.fit
## 1 2.882 0.071 0.8709432
## max.Sens.fit max.Spec.fit max.AUCROCR.fit opt.prob.threshold.fit
## 1 0.9632073 0.778679 0.8746354 0.6
## max.f.score.fit max.Accuracy.fit max.AccuracyLower.fit
## 1 0.8 0.9296422 0.9224771
## max.AccuracyUpper.fit max.Kappa.fit max.AUCpROC.OOB max.Sens.OOB
## 1 0.9371115 0.7515134 0.5870523 0.9045394
## max.Spec.OOB max.AUCROCR.OOB opt.prob.threshold.OOB max.f.score.OOB
## 1 0.2695652 0.5892132 0.6 0.2850575
## max.Accuracy.OOB max.AccuracyLower.OOB max.AccuracyUpper.OOB
## 1 0.8200231 0.8010821 0.8378705
## max.Kappa.OOB max.AccuracySD.fit max.KappaSD.fit
## 1 0.1825002 0.00506952 0.0191091if (!is.null(glb_date_vars) &&
(sum(grepl(paste(glb_date_vars, "\\.day\\.minutes\\.poly\\.", sep=""),
names(glbObsAll))) > 0)) {
# ret_lst <- myfit_mdl(mdl_id="Max.cor.Y.TmSrs.poly1",
# model_method=ifelse(glb_is_regression, "lm",
# ifelse(glb_is_binomial, "glm", "rpart")),
# model_type=glb_model_type,
# indep_vars_vctr=c(max_cor_y_x_vars, paste0(glb_date_vars, ".day.minutes")),
# rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
# fit_df=glbObsFit, OOB_df=glbObsOOB,
# n_cv_folds=glb_rcv_n_folds, tune_models_df=NULL)
#
ret_lst <- myfit_mdl(mdl_id="Max.cor.Y.TmSrs.poly",
model_method=ifelse(glb_is_regression, "lm",
ifelse(glb_is_binomial, "glm", "rpart")),
model_type=glb_model_type,
indep_vars_vctr=c(max_cor_y_x_vars,
grep(paste(glb_date_vars, "\\.day\\.minutes\\.poly\\.", sep=""),
names(glbObsAll), value=TRUE)),
rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
fit_df=glbObsFit, OOB_df=glbObsOOB,
n_cv_folds=glb_rcv_n_folds, tune_models_df=NULL)
}
# Interactions.High.cor.Y
if (length(int_feats <- setdiff(setdiff(unique(glb_feats_df$cor.high.X), NA),
subset(glb_feats_df, nzv)$id)) > 0) {
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix="Interact.High.cor.Y",
type=glb_model_type, trainControl.method="repeatedcv",
trainControl.number=glb_rcv_n_folds, trainControl.repeats=glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method="glmnet")),
indep_vars=c(max_cor_y_x_vars, paste(max_cor_y_x_vars[1], int_feats, sep=":")),
rsp_var=glb_rsp_var,
fit_df=glbObsFit, OOB_df=glbObsOOB)
} ## [1] "fitting model: Interact.High.cor.Y.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr,WordCount.root2:WordCount.root2"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0201 on full training set
## Length Class Mode
## a0 99 -none- numeric
## beta 2079 dgCMatrix S4
## df 99 -none- numeric
## dim 2 -none- numeric
## lambda 99 -none- numeric
## dev.ratio 99 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 21 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.89350373
## NDSSName.my.fctr#Multimedia#
## -0.01916344
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.18453357
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.21701058
## NDSSName.my.fctr#U.S.#Education
## -0.47679040
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.09891374
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.87281404
## NDSSName.my.fctrBusiness#Technology#
## 0.40965256
## NDSSName.my.fctrForeign#World#
## -0.05114617
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.47464340
## NDSSName.my.fctrOpEd#Opinion#
## 3.95214357
## NDSSName.my.fctrScience#Health#
## 3.14232408
## NDSSName.my.fctrStyles##Fashion
## -0.31867093
## NDSSName.my.fctrStyles#U.S.#
## 2.92610567
## NDSSName.my.fctrTStyle##
## -0.60538025
## WordCount.root2
## 0.05783392
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.95644632
## NDSSName.my.fctr#Multimedia#
## -0.07182859
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.30034382
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.29694236
## NDSSName.my.fctr#U.S.#Education
## -0.53415905
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.14259759
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.94231812
## NDSSName.my.fctrBusiness#Technology#
## 0.45657914
## NDSSName.my.fctrForeign#World#
## -0.10021084
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.53077048
## NDSSName.my.fctrOpEd#Opinion#
## 4.01324666
## NDSSName.my.fctrScience#Health#
## 3.18936803
## NDSSName.my.fctrStyles##Fashion
## -0.38069674
## NDSSName.my.fctrStyles#U.S.#
## 2.97176051
## NDSSName.my.fctrTStyle##
## -0.64837249
## WordCount.root2
## 0.05978318
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Interact.High.cor.Y.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Interact.High.cor.Y.glmnet.N
## 1 N 1146
## 2 Y 67
## Popular.fctr.predict.Interact.High.cor.Y.glmnet.Y
## 1 352
## 2 163
## Prediction
## Reference N Y
## N 1146 352
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.575231e-01 3.107477e-01 7.365992e-01 7.775689e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 9.066396e-44
## id
## 1 Interact.High.cor.Y.glmnet
## feats
## 1 WordCount.root2,NDSSName.my.fctr,WordCount.root2:WordCount.root2
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 4.395 0.27
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8767919 0.964476 0.7891078 0.9582555
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8099174 0.9333193
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7690803
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8067975
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4375839 0.7575231
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7365992 0.7775689 0.3107477
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.005178375 0.01754365# Low.cor.X
# if (glb_is_classification && glb_is_binomial)
# indep_vars_vctr <- subset(glb_feats_df, is.na(cor.high.X) &
# is.ConditionalX.y &
# (exclude.as.feat != 1))[, "id"] else
indep_vars <- subset(glb_feats_df, is.na(cor.high.X) & !nzv &
(exclude.as.feat != 1))[, "id"]
indep_vars <- myadjust_interaction_feats(indep_vars)
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix="Low.cor.X",
type=glb_model_type, trainControl.method="repeatedcv",
trainControl.number=glb_rcv_n_folds, trainControl.repeats=glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method="glmnet")),
indep_vars=indep_vars, rsp_var=glb_rsp_var,
fit_df=glbObsFit, OOB_df=glbObsOOB)## [1] "fitting model: Low.cor.X.glmnet"
## [1] " indep_vars: WordCount.root2,NDSSName.my.fctr,.rnorm,WordCount.nexp"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0201 on full training set
## Length Class Mode
## a0 100 -none- numeric
## beta 2300 dgCMatrix S4
## df 100 -none- numeric
## dim 2 -none- numeric
## lambda 100 -none- numeric
## dev.ratio 100 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 23 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -3.89350373
## NDSSName.my.fctr#Multimedia#
## -0.01916344
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.18453357
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.21701058
## NDSSName.my.fctr#U.S.#Education
## -0.47679040
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.09891374
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.87281404
## NDSSName.my.fctrBusiness#Technology#
## 0.40965256
## NDSSName.my.fctrForeign#World#
## -0.05114617
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.47464340
## NDSSName.my.fctrOpEd#Opinion#
## 3.95214357
## NDSSName.my.fctrScience#Health#
## 3.14232408
## NDSSName.my.fctrStyles##Fashion
## -0.31867093
## NDSSName.my.fctrStyles#U.S.#
## 2.92610567
## NDSSName.my.fctrTStyle##
## -0.60538025
## WordCount.root2
## 0.05783392
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -3.95644632
## NDSSName.my.fctr#Multimedia#
## -0.07182859
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.30034382
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.29694236
## NDSSName.my.fctr#U.S.#Education
## -0.53415905
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.14259759
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.94231812
## NDSSName.my.fctrBusiness#Technology#
## 0.45657914
## NDSSName.my.fctrForeign#World#
## -0.10021084
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.53077048
## NDSSName.my.fctrOpEd#Opinion#
## 4.01324666
## NDSSName.my.fctrScience#Health#
## 3.18936803
## NDSSName.my.fctrStyles##Fashion
## -0.38069674
## NDSSName.my.fctrStyles#U.S.#
## 2.97176051
## NDSSName.my.fctrTStyle##
## -0.64837249
## WordCount.root2
## 0.05978318
## Popular.fctr Popular.fctr.predict.Low.cor.X.glmnet.N
## 1 N 3796
## 2 Y 177
## Popular.fctr.predict.Low.cor.X.glmnet.Y
## 1 145
## 2 686
## Prediction
## Reference N Y
## N 3796 145
## Y 177 686
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.329725e-01 7.692476e-01 9.255302e-01 9.398832e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 1.026390e-114 8.406670e-02
## Popular.fctr Popular.fctr.predict.Low.cor.X.glmnet.N
## 1 N 1146
## 2 Y 67
## Popular.fctr.predict.Low.cor.X.glmnet.Y
## 1 352
## 2 163
## Prediction
## Reference N Y
## N 1146 352
## Y 67 163
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.575231e-01 3.107477e-01 7.365992e-01 7.775689e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 9.066396e-44
## id feats
## 1 Low.cor.X.glmnet WordCount.root2,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 4.549 0.293
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8767919 0.964476 0.7891078 0.9582555
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8099174 0.9333193
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9255302 0.9398832 0.7690803
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.5962443 0.9098798 0.2826087 0.8067975
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4375839 0.7575231
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.7365992 0.7775689 0.3107477
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.005178375 0.01754365rm(ret_lst)
glb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.models", major.inc=FALSE)## label step_major step_minor label_minor bgn end elapsed
## 10 fit.models 6 0 0 35.238 139.59 104.352
## 11 fit.models 6 1 1 139.591 NA NAfit.models_1_chunk_df <- myadd_chunk(NULL, "fit.models_1_bgn", label.minor="setup")## label step_major step_minor label_minor bgn end elapsed
## 1 fit.models_1_bgn 1 0 setup 151.043 NA NA#stop(here"); glb_to_sav(); all.equal(glb_models_df, sav_models_df)
topindep_var <- NULL; interact_vars <- NULL;
for (mdl_id_pfx in names(glb_mdl_family_lst)) {
fit.models_1_chunk_df <-
myadd_chunk(fit.models_1_chunk_df, paste0("fit.models_1_", mdl_id_pfx),
major.inc = FALSE, label.minor = "setup")
indep_vars <- NULL;
if (grepl("\\.Interact", mdl_id_pfx)) {
if (is.null(topindep_var) && is.null(interact_vars)) {
# select best glmnet model upto now
dsp_models_df <- orderBy(model_sel_frmla <- get_model_sel_frmla(),
glb_models_df)
dsp_models_df <- subset(dsp_models_df,
grepl(".glmnet", id, fixed = TRUE))
bst_mdl_id <- dsp_models_df$id[1]
mdl_id_pfx <-
paste(c(head(unlist(strsplit(bst_mdl_id, "[.]")), -1), "Interact"),
collapse=".")
# select most importance feature
if (is.null(bst_featsimp_df <-
myget_feats_importance(glb_models_lst[[bst_mdl_id]]))) {
warning("Base model for RFE.Interact: ", bst_mdl_id,
" has no important features")
next
}
topindep_ix <- 1
while (is.null(topindep_var) && (topindep_ix <= nrow(bst_featsimp_df))) {
topindep_var <- row.names(bst_featsimp_df)[topindep_ix]
if (grepl(".fctr", topindep_var, fixed=TRUE))
topindep_var <-
paste0(unlist(strsplit(topindep_var, ".fctr"))[1], ".fctr")
if (topindep_var %in% names(glbFeatsInteractionOnly)) {
topindep_var <- NULL; topindep_ix <- topindep_ix + 1
} else break
}
# select features with importance > max(10, importance of .rnorm) & is not highest
# combine factor dummy features to just the factor feature
if (length(pos_rnorm <-
grep(".rnorm", row.names(bst_featsimp_df), fixed=TRUE)) > 0)
imp_rnorm <- bst_featsimp_df[pos_rnorm, 1] else
imp_rnorm <- NA
importance_cutoff <- max(10, imp_rnorm, na.rm=TRUE)
interact_vars <-
tail(row.names(subset(bst_featsimp_df,
importance > importance_cutoff)), -1)
if (length(interact_vars) > 0) {
interact_vars <-
myadjust_interaction_feats(myextract_actual_feats(interact_vars))
interact_vars <-
interact_vars[!grepl(topindep_var, interact_vars, fixed=TRUE)]
}
### bid0_sp only
# interact_vars <- c(
# "biddable", "D.ratio.sum.TfIdf.wrds.n", "D.TfIdf.sum.stem.stop.Ratio", "D.sum.TfIdf",
# "D.TfIdf.sum.post.stop", "D.TfIdf.sum.post.stem", "D.ratio.wrds.stop.n.wrds.n", "D.chrs.uppr.n.log",
# "D.chrs.n.log", "color.fctr"
# # , "condition.fctr", "prdl.my.descr.fctr"
# )
# interact_vars <- setdiff(interact_vars, c("startprice.dgt2.is9", "color.fctr"))
###
indep_vars <- myextract_actual_feats(row.names(bst_featsimp_df))
indep_vars <- setdiff(indep_vars, topindep_var)
if (length(interact_vars) > 0) {
indep_vars <-
setdiff(indep_vars, myextract_actual_feats(interact_vars))
indep_vars <- c(indep_vars,
paste(topindep_var, setdiff(interact_vars, topindep_var),
sep = "*"))
} else indep_vars <- union(indep_vars, topindep_var)
}
}
if (is.null(indep_vars))
indep_vars <- glb_mdl_feats_lst[[mdl_id_pfx]]
if (is.null(indep_vars) && grepl("RFE\\.", mdl_id_pfx))
indep_vars <- myextract_actual_feats(predictors(rfe_fit_results))
if (is.null(indep_vars))
indep_vars <- subset(glb_feats_df, !nzv & (exclude.as.feat != 1))[, "id"]
if (grepl("^%<d-%", indep_vars)) {
#stop(here")
indep_vars <-
eval(parse(text = str_trim(unlist(strsplit(indep_vars, "%<d-%"))[2])))
}
indep_vars <- myadjust_interaction_feats(indep_vars)
if (grepl("\\.Interact", mdl_id_pfx)) {
# if (method != tail(unlist(strsplit(bst_mdl_id, "[.]")), 1)) next
if (is.null(glb_mdl_family_lst[[mdl_id_pfx]])) {
if (!is.null(glb_mdl_family_lst[["Best.Interact"]]))
glb_mdl_family_lst[[mdl_id_pfx]] <-
glb_mdl_family_lst[["Best.Interact"]]
}
}
if (!is.null(glbObsFitOutliers[[mdl_id_pfx]])) {
fitobs_df <- glbObsFit[!(glbObsFit[, glb_id_var] %in%
glbObsFitOutliers[[mdl_id_pfx]]), ]
} else fitobs_df <- glbObsFit
if (is.null(glb_mdl_family_lst[[mdl_id_pfx]]))
mdl_methods <- glbMdlMethods else
mdl_methods <- glb_mdl_family_lst[[mdl_id_pfx]]
for (method in mdl_methods) {
if (method %in% c("rpart", "rf")) {
# rpart: fubar's the tree
# rf: skip the scenario w/ .rnorm for speed
indep_vars <- setdiff(indep_vars, c(".rnorm"))
#mdl_id <- paste0(mdl_id_pfx, ".no.rnorm")
}
fit.models_1_chunk_df <- myadd_chunk(fit.models_1_chunk_df,
paste0("fit.models_1_", mdl_id_pfx), major.inc = FALSE,
label.minor = method)
ret_lst <-
myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = mdl_id_pfx,
type = glb_model_type, tune.df = glb_tune_models_df,
trainControl.method = "repeatedcv",
trainControl.number = glb_rcv_n_folds,
trainControl.repeats = glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
#trainControl.allowParallel = FALSE,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method = method)),
indep_vars = indep_vars, rsp_var = glb_rsp_var,
fit_df = fitobs_df, OOB_df = glbObsOOB)
}
}## label step_major step_minor label_minor bgn end
## 1 fit.models_1_bgn 1 0 setup 151.043 151.052
## 2 fit.models_1_All.X 1 1 setup 151.052 NA
## elapsed
## 1 0.009
## 2 NA## Warning in if (grepl("^%<d-%", indep_vars)) {: the condition has length > 1
## and only the first element will be used## label step_major step_minor label_minor bgn end
## 2 fit.models_1_All.X 1 1 setup 151.052 151.059
## 3 fit.models_1_All.X 1 2 glmnet 151.060 NA
## elapsed
## 2 0.007
## 3 NA
## [1] "fitting model: All.X.glmnet"
## [1] " indep_vars: WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp"
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 1, lambda = 0.00434 on full training set## Warning in myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst
## = list(id.prefix = mdl_id_pfx, : model's bestTune found at an extreme of
## tuneGrid for parameter: alpha
## Length Class Mode
## a0 91 -none- numeric
## beta 2184 dgCMatrix S4
## df 91 -none- numeric
## dim 2 -none- numeric
## lambda 91 -none- numeric
## dev.ratio 91 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 24 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -5.52332827
## NDSSName.my.fctr#Multimedia#
## -0.16474095
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.18132553
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.90787932
## NDSSName.my.fctr#U.S.#Education
## -0.90764328
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.24502067
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.67941504
## NDSSName.my.fctrBusiness#Technology#
## 0.83419892
## NDSSName.my.fctrForeign#World#
## -0.08805297
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.83659624
## NDSSName.my.fctrOpEd#Opinion#
## 4.79265681
## NDSSName.my.fctrScience#Health#
## 3.67959505
## NDSSName.my.fctrStyles##Fashion
## -0.64461547
## NDSSName.my.fctrStyles#U.S.#
## 3.44343706
## NDSSName.my.fctrTStyle##
## -0.97236910
## NDSSName.my.fctrmyOther
## -0.23762498
## WordCount.log1p
## 0.18619880
## WordCount.root2
## 0.06691104
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -5.59881022
## NDSSName.my.fctr#Multimedia#
## -0.24609678
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.30818954
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.94273791
## NDSSName.my.fctr#U.S.#Education
## -1.01466873
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.30479819
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.69839363
## NDSSName.my.fctrBusiness#Technology#
## 0.84455935
## NDSSName.my.fctrForeign#World#
## -0.18204983
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.91223472
## NDSSName.my.fctrOpEd#Opinion#
## 4.80649138
## NDSSName.my.fctrScience#Health#
## 3.68561250
## NDSSName.my.fctrStyles##Fashion
## -0.74721366
## NDSSName.my.fctrStyles#U.S.#
## 3.44963924
## NDSSName.my.fctrTStyle##
## -1.04099014
## NDSSName.my.fctrTravel#Travel#
## -0.03203556
## NDSSName.my.fctrmyOther
## -0.35136209
## WordCount.log1p
## 0.19889748
## WordCount.root2
## 0.06700663
## Popular.fctr Popular.fctr.predict.All.X.glmnet.N
## 1 N 3793
## 2 Y 176
## Popular.fctr.predict.All.X.glmnet.Y
## 1 148
## 2 687
## Prediction
## Reference N Y
## N 3793 148
## Y 176 687
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.325562e-01 7.682400e-01 9.250938e-01 9.394875e-01 8.203580e-01
## AccuracyPValue McnemarPValue
## 9.476102e-114 1.336144e-01
## Popular.fctr Popular.fctr.predict.All.X.glmnet.N
## 1 N 1176
## 2 Y 77
## Popular.fctr.predict.All.X.glmnet.Y
## 1 322
## 2 153
## Prediction
## Reference N Y
## N 1176 322
## Y 77 153
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 7.690972e-01 3.103487e-01 7.484910e-01 7.887855e-01 8.668981e-01
## AccuracyPValue McnemarPValue
## 1.000000e+00 2.576173e-34
## id
## 1 All.X.glmnet
## feats
## 1 WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 4.817 0.294
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.8771175 0.9639685 0.7902665 0.9589072
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.8091873 0.9326952
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.9250938 0.9394875 0.7668828
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB max.AUCROCR.OOB
## 1 0.596578 0.9105474 0.2826087 0.8061009
## opt.prob.threshold.OOB max.f.score.OOB max.Accuracy.OOB
## 1 0.1 0.4340426 0.7690972
## max.AccuracyLower.OOB max.AccuracyUpper.OOB max.Kappa.OOB
## 1 0.748491 0.7887855 0.3103487
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.004892081 0.01659075# Check if other preProcess methods improve model performance
fit.models_1_chunk_df <-
myadd_chunk(fit.models_1_chunk_df, "fit.models_1_preProc", major.inc = FALSE,
label.minor = "preProc")## label step_major step_minor label_minor bgn end
## 3 fit.models_1_All.X 1 2 glmnet 151.060 161.253
## 4 fit.models_1_preProc 1 3 preProc 161.253 NA
## elapsed
## 3 10.193
## 4 NAmdl_id <- orderBy(get_model_sel_frmla(), glb_models_df)[1, "id"]
indep_vars_vctr <- trim(unlist(strsplit(glb_models_df[glb_models_df$id == mdl_id,
"feats"], "[,]")))
method <- tail(unlist(strsplit(mdl_id, "[.]")), 1)
mdl_id_pfx <- paste0(head(unlist(strsplit(mdl_id, "[.]")), -1), collapse = ".")
if (!is.null(glbObsFitOutliers[[mdl_id_pfx]])) {
fitobs_df <- glbObsFit[!(glbObsFit[, glb_id_var] %in%
glbObsFitOutliers[[mdl_id_pfx]]), ]
} else fitobs_df <- glbObsFit
for (prePr in glb_preproc_methods) {
# The operations are applied in this order:
# Box-Cox/Yeo-Johnson transformation, centering, scaling, range, imputation, PCA, ICA then spatial sign.
ret_lst <- myfit_mdl(mdl_specs_lst=myinit_mdl_specs_lst(mdl_specs_lst=list(
id.prefix=mdl_id_pfx,
type=glb_model_type, tune.df=glb_tune_models_df,
trainControl.method="repeatedcv",
trainControl.number=glb_rcv_n_folds,
trainControl.repeats=glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method=method, train.preProcess=prePr)),
indep_vars=indep_vars_vctr, rsp_var=glb_rsp_var,
fit_df=fitobs_df, OOB_df=glbObsOOB)
}
# If (All|RFE).X.glm is less accurate than Low.Cor.X.glm
# check NA coefficients & filter appropriate terms in indep_vars_vctr
# if (method == "glm") {
# orig_glm <- glb_models_lst[[paste0(mdl_id, ".", model_method)]]$finalModel
# orig_glm <- glb_models_lst[["All.X.glm"]]$finalModel; print(summary(orig_glm))
# orig_glm <- glb_models_lst[["RFE.X.glm"]]$finalModel; print(summary(orig_glm))
# require(car)
# vif_orig_glm <- vif(orig_glm); print(vif_orig_glm)
# # if vif errors out with "there are aliased coefficients in the model"
# alias_orig_glm <- alias(orig_glm); alias_complete_orig_glm <- (alias_orig_glm$Complete > 0); alias_complete_orig_glm <- alias_complete_orig_glm[rowSums(alias_complete_orig_glm) > 0, colSums(alias_complete_orig_glm) > 0]; print(alias_complete_orig_glm)
# print(vif_orig_glm[!is.na(vif_orig_glm) & (vif_orig_glm == Inf)])
# print(which.max(vif_orig_glm))
# print(sort(vif_orig_glm[vif_orig_glm >= 1.0e+03], decreasing=TRUE))
# glbObsFit[c(1143, 3637, 3953, 4105), c("UniqueID", "Popular", "H.P.quandary", "Headline")]
# glb_feats_df[glb_feats_df$id %in% grep("[HSA]\\.chrs.n.log", glb_feats_df$id, value=TRUE) | glb_feats_df$cor.high.X %in% grep("[HSA]\\.chrs.n.log", glb_feats_df$id, value=TRUE), ]
# all.equal(glbObsAll$S.chrs.uppr.n.log, glbObsAll$A.chrs.uppr.n.log)
# cor(glbObsAll$S.T.herald, glbObsAll$S.T.tribun)
# mydspObs(Abstract.contains="[Dd]iar", cols=("Abstract"), all=TRUE)
# subset(glb_feats_df, cor.y.abs <= glb_feats_df[glb_feats_df$id == ".rnorm", "cor.y.abs"])
# corxx_mtrx <- cor(data.matrix(glbObsAll[, setdiff(names(glbObsAll), myfind_chr_cols_df(glbObsAll))]), use="pairwise.complete.obs"); abs_corxx_mtrx <- abs(corxx_mtrx); diag(abs_corxx_mtrx) <- 0
# which.max(abs_corxx_mtrx["S.T.tribun", ])
# abs_corxx_mtrx["A.npnct08.log", "S.npnct08.log"]
# step_glm <- step(orig_glm)
# }
# Since caret does not optimize rpart well
# if (method == "rpart")
# ret_lst <- myfit_mdl(mdl_id=paste0(mdl_id_pfx, ".cp.0"), model_method=method,
# indep_vars_vctr=indep_vars_vctr,
# model_type=glb_model_type,
# rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
# fit_df=glbObsFit, OOB_df=glbObsOOB,
# n_cv_folds=0, tune_models_df=data.frame(parameter="cp", min=0.0, max=0.0, by=0.1))
# User specified
# Ensure at least 2 vars in each regression; else varImp crashes
# sav_models_lst <- glb_models_lst; sav_models_df <- glb_models_df; sav_featsimp_df <- glb_featsimp_df; all.equal(sav_featsimp_df, glb_featsimp_df)
# glb_models_lst <- sav_models_lst; glb_models_df <- sav_models_df; glm_featsimp_df <- sav_featsimp_df
# easier to exclude features
# require(gdata) # needed for trim
# mdl_id <- "";
# indep_vars_vctr <- head(subset(glb_models_df, grepl("All\\.X\\.", mdl_id), select=feats)
# , 1)[, "feats"]
# indep_vars_vctr <- trim(unlist(strsplit(indep_vars_vctr, "[,]")))
# indep_vars_vctr <- setdiff(indep_vars_vctr, ".rnorm")
# easier to include features
#stop(here"); sav_models_df <- glb_models_df; glb_models_df <- sav_models_df
# !_sp
# mdl_id <- "csm"; indep_vars_vctr <- c(NULL
# ,"prdline.my.fctr", "prdline.my.fctr:.clusterid.fctr"
# ,"prdline.my.fctr*biddable"
# #,"prdline.my.fctr*startprice.log"
# #,"prdline.my.fctr*startprice.diff"
# ,"prdline.my.fctr*condition.fctr"
# ,"prdline.my.fctr*D.terms.post.stop.n"
# #,"prdline.my.fctr*D.terms.post.stem.n"
# ,"prdline.my.fctr*cellular.fctr"
# # ,"<feat1>:<feat2>"
# )
# for (method in glbMdlMethods) {
# ret_lst <- myfit_mdl(mdl_id=mdl_id, model_method=method,
# indep_vars_vctr=indep_vars_vctr,
# model_type=glb_model_type,
# rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
# fit_df=glbObsFit, OOB_df=glbObsOOB,
# n_cv_folds=glb_rcv_n_folds, tune_models_df=glb_tune_models_df)
# csm_mdl_id <- paste0(mdl_id, ".", method)
# csm_featsimp_df <- myget_feats_importance(glb_models_lst[[paste0(mdl_id, ".",
# method)]]); print(head(csm_featsimp_df))
# }
###
# Ntv.1.lm <- lm(reformulate(indep_vars_vctr, glb_rsp_var), glbObsTrn); print(summary(Ntv.1.lm))
#glb_models_df[, "max.Accuracy.OOB", FALSE]
#varImp(glb_models_lst[["Low.cor.X.glm"]])
#orderBy(~ -Overall, varImp(glb_models_lst[["All.X.2.glm"]])$importance)
#orderBy(~ -Overall, varImp(glb_models_lst[["All.X.3.glm"]])$importance)
#glb_feats_df[grepl("npnct28", glb_feats_df$id), ]
# User specified bivariate models
# indep_vars_vctr_lst <- list()
# for (feat in setdiff(names(glbObsFit),
# union(glb_rsp_var, glbFeatsExclude)))
# indep_vars_vctr_lst[["feat"]] <- feat
# User specified combinatorial models
# indep_vars_vctr_lst <- list()
# combn_mtrx <- combn(c("<feat1_name>", "<feat2_name>", "<featn_name>"),
# <num_feats_to_choose>)
# for (combn_ix in 1:ncol(combn_mtrx))
# #print(combn_mtrx[, combn_ix])
# indep_vars_vctr_lst[[combn_ix]] <- combn_mtrx[, combn_ix]
# template for myfit_mdl
# rf is hard-coded in caret to recognize only Accuracy / Kappa evaluation metrics
# only for OOB in trainControl ?
# ret_lst <- myfit_mdl_fn(mdl_id=paste0(mdl_id_pfx, ""), model_method=method,
# indep_vars_vctr=indep_vars_vctr,
# rsp_var=glb_rsp_var, rsp_var_out=glb_rsp_var_out,
# fit_df=glbObsFit, OOB_df=glbObsOOB,
# n_cv_folds=glb_rcv_n_folds, tune_models_df=glb_tune_models_df,
# model_loss_mtrx=glbMdlMetric_terms,
# model_summaryFunction=glbMdlMetricSummaryFn,
# model_metric=glbMdlMetricSummary,
# model_metric_maximize=glbMdlMetricMaximize)
# Simplify a model
# fit_df <- glbObsFit; glb_mdl <- step(<complex>_mdl)
# Non-caret models
# rpart_area_mdl <- rpart(reformulate("Area", response=glb_rsp_var),
# data=glbObsFit, #method="class",
# control=rpart.control(cp=0.12),
# parms=list(loss=glbMdlMetric_terms))
# print("rpart_sel_wlm_mdl"); prp(rpart_sel_wlm_mdl)
#
print(glb_models_df)## id
## MFO.myMFO_classfr MFO.myMFO_classfr
## Random.myrandom_classfr Random.myrandom_classfr
## Max.cor.Y.rcv.1X1.glmnet Max.cor.Y.rcv.1X1.glmnet
## Max.cor.Y.rcv.3X1.glmnet Max.cor.Y.rcv.3X1.glmnet
## Max.cor.Y.rcv.3X3.glmnet Max.cor.Y.rcv.3X3.glmnet
## Max.cor.Y.rcv.3X5.glmnet Max.cor.Y.rcv.3X5.glmnet
## Max.cor.Y.rcv.5X1.glmnet Max.cor.Y.rcv.5X1.glmnet
## Max.cor.Y.rcv.5X3.glmnet Max.cor.Y.rcv.5X3.glmnet
## Max.cor.Y.rcv.5X5.glmnet Max.cor.Y.rcv.5X5.glmnet
## Max.cor.Y.rcv.1X1.cp.0.rpart Max.cor.Y.rcv.1X1.cp.0.rpart
## Max.cor.Y.rpart Max.cor.Y.rpart
## Interact.High.cor.Y.glmnet Interact.High.cor.Y.glmnet
## Low.cor.X.glmnet Low.cor.X.glmnet
## All.X.glmnet All.X.glmnet
## feats
## MFO.myMFO_classfr .rnorm
## Random.myrandom_classfr .rnorm
## Max.cor.Y.rcv.1X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X3.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X5.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X3.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X5.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.1X1.cp.0.rpart WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rpart WordCount.root2,NDSSName.my.fctr
## Interact.High.cor.Y.glmnet WordCount.root2,NDSSName.my.fctr,WordCount.root2:WordCount.root2
## Low.cor.X.glmnet WordCount.root2,NDSSName.my.fctr,.rnorm,WordCount.nexp
## All.X.glmnet WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns min.elapsedtime.everything
## MFO.myMFO_classfr 0 0.269
## Random.myrandom_classfr 0 0.269
## Max.cor.Y.rcv.1X1.glmnet 0 0.972
## Max.cor.Y.rcv.3X1.glmnet 25 2.826
## Max.cor.Y.rcv.3X3.glmnet 25 4.817
## Max.cor.Y.rcv.3X5.glmnet 25 6.419
## Max.cor.Y.rcv.5X1.glmnet 25 3.223
## Max.cor.Y.rcv.5X3.glmnet 25 6.142
## Max.cor.Y.rcv.5X5.glmnet 25 8.908
## Max.cor.Y.rcv.1X1.cp.0.rpart 0 0.863
## Max.cor.Y.rpart 5 2.882
## Interact.High.cor.Y.glmnet 25 4.395
## Low.cor.X.glmnet 25 4.549
## All.X.glmnet 25 4.817
## min.elapsedtime.final max.AUCpROC.fit
## MFO.myMFO_classfr 0.003 0.5000000
## Random.myrandom_classfr 0.001 0.4990604
## Max.cor.Y.rcv.1X1.glmnet 0.273 0.8790544
## Max.cor.Y.rcv.3X1.glmnet 0.270 0.8767919
## Max.cor.Y.rcv.3X3.glmnet 0.270 0.8767919
## Max.cor.Y.rcv.3X5.glmnet 0.269 0.8767919
## Max.cor.Y.rcv.5X1.glmnet 0.267 0.8784031
## Max.cor.Y.rcv.5X3.glmnet 0.267 0.8784031
## Max.cor.Y.rcv.5X5.glmnet 0.268 0.8784031
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.070 0.8821543
## Max.cor.Y.rpart 0.071 0.8709432
## Interact.High.cor.Y.glmnet 0.270 0.8767919
## Low.cor.X.glmnet 0.293 0.8767919
## All.X.glmnet 0.294 0.8771175
## max.Sens.fit max.Spec.fit max.AUCROCR.fit
## MFO.myMFO_classfr 1.0000000 0.0000000 0.5000000
## Random.myrandom_classfr 0.8312611 0.1668598 0.4972757
## Max.cor.Y.rcv.1X1.glmnet 0.9632073 0.7949015 0.9608594
## Max.cor.Y.rcv.3X1.glmnet 0.9644760 0.7891078 0.9582555
## Max.cor.Y.rcv.3X3.glmnet 0.9644760 0.7891078 0.9582555
## Max.cor.Y.rcv.3X5.glmnet 0.9644760 0.7891078 0.9582555
## Max.cor.Y.rcv.5X1.glmnet 0.9642223 0.7925840 0.9607052
## Max.cor.Y.rcv.5X3.glmnet 0.9642223 0.7925840 0.9607052
## Max.cor.Y.rcv.5X5.glmnet 0.9642223 0.7925840 0.9607052
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9705658 0.7937428 0.9504198
## Max.cor.Y.rpart 0.9632073 0.7786790 0.8746354
## Interact.High.cor.Y.glmnet 0.9644760 0.7891078 0.9582555
## Low.cor.X.glmnet 0.9644760 0.7891078 0.9582555
## All.X.glmnet 0.9639685 0.7902665 0.9589072
## opt.prob.threshold.fit max.f.score.fit
## MFO.myMFO_classfr 0.1 0.3045703
## Random.myrandom_classfr 0.1 0.3045703
## Max.cor.Y.rcv.1X1.glmnet 0.5 0.8099174
## Max.cor.Y.rcv.3X1.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.3X3.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.3X5.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.5X1.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.5X3.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.5X5.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.4 0.8235294
## Max.cor.Y.rpart 0.6 0.8000000
## Interact.High.cor.Y.glmnet 0.4 0.8099174
## Low.cor.X.glmnet 0.4 0.8099174
## All.X.glmnet 0.4 0.8091873
## max.Accuracy.fit max.AccuracyLower.fit
## MFO.myMFO_classfr 0.1796420 0.1688795
## Random.myrandom_classfr 0.1796420 0.1688795
## Max.cor.Y.rcv.1X1.glmnet 0.9329725 0.9255302
## Max.cor.Y.rcv.3X1.glmnet 0.9335973 0.9255302
## Max.cor.Y.rcv.3X3.glmnet 0.9333193 0.9255302
## Max.cor.Y.rcv.3X5.glmnet 0.9332218 0.9255302
## Max.cor.Y.rcv.5X1.glmnet 0.9331818 0.9259666
## Max.cor.Y.rcv.5X3.glmnet 0.9333905 0.9259666
## Max.cor.Y.rcv.5X5.glmnet 0.9331816 0.9259666
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9381765 0.9309917
## Max.cor.Y.rpart 0.9296422 0.9224771
## Interact.High.cor.Y.glmnet 0.9333193 0.9255302
## Low.cor.X.glmnet 0.9333193 0.9255302
## All.X.glmnet 0.9326952 0.9250938
## max.AccuracyUpper.fit max.Kappa.fit
## MFO.myMFO_classfr 0.1907952 0.0000000
## Random.myrandom_classfr 0.1907952 0.0000000
## Max.cor.Y.rcv.1X1.glmnet 0.9398832 0.7692476
## Max.cor.Y.rcv.3X1.glmnet 0.9398832 0.7691678
## Max.cor.Y.rcv.3X3.glmnet 0.9398832 0.7690803
## Max.cor.Y.rcv.3X5.glmnet 0.9398832 0.7686375
## Max.cor.Y.rcv.5X1.glmnet 0.9402789 0.7689055
## Max.cor.Y.rcv.5X3.glmnet 0.9402789 0.7698577
## Max.cor.Y.rcv.5X5.glmnet 0.9402789 0.7691429
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9448229 0.7860827
## Max.cor.Y.rpart 0.9371115 0.7515134
## Interact.High.cor.Y.glmnet 0.9398832 0.7690803
## Low.cor.X.glmnet 0.9398832 0.7690803
## All.X.glmnet 0.9394875 0.7668828
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB
## MFO.myMFO_classfr 0.5000000 1.0000000 0.0000000
## Random.myrandom_classfr 0.5125675 0.8077437 0.2173913
## Max.cor.Y.rcv.1X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X3.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X5.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X3.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X5.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.6174697 0.9218959 0.3130435
## Max.cor.Y.rpart 0.5870523 0.9045394 0.2695652
## Interact.High.cor.Y.glmnet 0.5962443 0.9098798 0.2826087
## Low.cor.X.glmnet 0.5962443 0.9098798 0.2826087
## All.X.glmnet 0.5965780 0.9105474 0.2826087
## max.AUCROCR.OOB opt.prob.threshold.OOB
## MFO.myMFO_classfr 0.5000000 0.1
## Random.myrandom_classfr 0.4857956 0.1
## Max.cor.Y.rcv.1X1.glmnet 0.8116126 0.1
## Max.cor.Y.rcv.3X1.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.3X3.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.3X5.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.5X1.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.5X3.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.5X5.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7773858 0.1
## Max.cor.Y.rpart 0.5892132 0.6
## Interact.High.cor.Y.glmnet 0.8067975 0.1
## Low.cor.X.glmnet 0.8067975 0.1
## All.X.glmnet 0.8061009 0.1
## max.f.score.OOB max.Accuracy.OOB
## MFO.myMFO_classfr 0.2349336 0.1331019
## Random.myrandom_classfr 0.2349336 0.1331019
## Max.cor.Y.rcv.1X1.glmnet 0.4405405 0.7604167
## Max.cor.Y.rcv.3X1.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.3X3.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.3X5.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.5X1.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.5X3.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.5X5.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.4207493 0.7673611
## Max.cor.Y.rpart 0.2850575 0.8200231
## Interact.High.cor.Y.glmnet 0.4375839 0.7575231
## Low.cor.X.glmnet 0.4375839 0.7575231
## All.X.glmnet 0.4340426 0.7690972
## max.AccuracyLower.OOB max.AccuracyUpper.OOB
## MFO.myMFO_classfr 0.1174298 0.1500310
## Random.myrandom_classfr 0.1174298 0.1500310
## Max.cor.Y.rcv.1X1.glmnet 0.7395703 0.7803749
## Max.cor.Y.rcv.3X1.glmnet 0.7365992 0.7775689
## Max.cor.Y.rcv.3X3.glmnet 0.7365992 0.7775689
## Max.cor.Y.rcv.3X5.glmnet 0.7365992 0.7775689
## Max.cor.Y.rcv.5X1.glmnet 0.7395703 0.7803749
## Max.cor.Y.rcv.5X3.glmnet 0.7395703 0.7803749
## Max.cor.Y.rcv.5X5.glmnet 0.7395703 0.7803749
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7467059 0.7871043
## Max.cor.Y.rpart 0.8010821 0.8378705
## Interact.High.cor.Y.glmnet 0.7365992 0.7775689
## Low.cor.X.glmnet 0.7365992 0.7775689
## All.X.glmnet 0.7484910 0.7887855
## max.Kappa.OOB max.AccuracySD.fit
## MFO.myMFO_classfr 0.0000000 NA
## Random.myrandom_classfr 0.0000000 NA
## Max.cor.Y.rcv.1X1.glmnet 0.3148374 NA
## Max.cor.Y.rcv.3X1.glmnet 0.3107477 0.007015493
## Max.cor.Y.rcv.3X3.glmnet 0.3107477 0.005178375
## Max.cor.Y.rcv.3X5.glmnet 0.3107477 0.005396525
## Max.cor.Y.rcv.5X1.glmnet 0.3373693 0.008837283
## Max.cor.Y.rcv.5X3.glmnet 0.3373693 0.006138477
## Max.cor.Y.rcv.5X5.glmnet 0.3373693 0.006213800
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.2953321 NA
## Max.cor.Y.rpart 0.1825002 0.005069520
## Interact.High.cor.Y.glmnet 0.3107477 0.005178375
## Low.cor.X.glmnet 0.3107477 0.005178375
## All.X.glmnet 0.3103487 0.004892081
## max.KappaSD.fit
## MFO.myMFO_classfr NA
## Random.myrandom_classfr NA
## Max.cor.Y.rcv.1X1.glmnet NA
## Max.cor.Y.rcv.3X1.glmnet 0.02403706
## Max.cor.Y.rcv.3X3.glmnet 0.01754365
## Max.cor.Y.rcv.3X5.glmnet 0.01835474
## Max.cor.Y.rcv.5X1.glmnet 0.03133449
## Max.cor.Y.rcv.5X3.glmnet 0.02161286
## Max.cor.Y.rcv.5X5.glmnet 0.02210061
## Max.cor.Y.rcv.1X1.cp.0.rpart NA
## Max.cor.Y.rpart 0.01910910
## Interact.High.cor.Y.glmnet 0.01754365
## Low.cor.X.glmnet 0.01754365
## All.X.glmnet 0.01659075rm(ret_lst)
fit.models_1_chunk_df <-
myadd_chunk(fit.models_1_chunk_df, "fit.models_1_end", major.inc = FALSE,
label.minor = "teardown")## label step_major step_minor label_minor bgn end
## 4 fit.models_1_preProc 1 3 preProc 161.253 161.32
## 5 fit.models_1_end 1 4 teardown 161.320 NA
## elapsed
## 4 0.067
## 5 NAglb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.models", major.inc = FALSE)## label step_major step_minor label_minor bgn end elapsed
## 11 fit.models 6 1 1 139.591 161.329 21.738
## 12 fit.models 6 2 2 161.330 NA NAfit.models_2_chunk_df <-
myadd_chunk(NULL, "fit.models_2_bgn", label.minor = "setup")## label step_major step_minor label_minor bgn end elapsed
## 1 fit.models_2_bgn 1 0 setup 163.004 NA NA#stop(here"); glb_to_sav(); all.equal(glb_models_df, sav_models_df)
# if (!is.null(glbMdlMetricSummaryFn)) {
# stats_df <- glb_models_df[, "id", FALSE]
#
# stats_mdl_df <- data.frame()
# for (mdl_id in stats_df$id) {
# stats_mdl_df <- rbind(stats_mdl_df,
# mypredict_mdl(glb_models_lst[[mdl_id]], glbObsFit, glb_rsp_var,
# glb_rsp_var_out, mdl_id, "fit",
# glbMdlMetricSummaryFn, glbMdlMetricSummary,
# glbMdlMetricMaximize, ret_type="stats"))
# }
# stats_df <- merge(stats_df, stats_mdl_df, all.x=TRUE)
#
# stats_mdl_df <- data.frame()
# for (mdl_id in stats_df$id) {
# stats_mdl_df <- rbind(stats_mdl_df,
# mypredict_mdl(glb_models_lst[[mdl_id]], glbObsOOB, glb_rsp_var,
# glb_rsp_var_out, mdl_id, "OOB",
# glbMdlMetricSummaryFn, glbMdlMetricSummary,
# glbMdlMetricMaximize, ret_type="stats"))
# }
# stats_df <- merge(stats_df, stats_mdl_df, all.x=TRUE)
#
# print("Merging following data into glb_models_df:")
# print(stats_mrg_df <- stats_df[, c(1, grep(glbMdlMetricSummary, names(stats_df)))])
# print(tmp_models_df <- orderBy(~id, glb_models_df[, c("id",
# grep(glbMdlMetricSummary, names(stats_df), value=TRUE))]))
#
# tmp2_models_df <- glb_models_df[, c("id", setdiff(names(glb_models_df),
# grep(glbMdlMetricSummary, names(stats_df), value=TRUE)))]
# tmp3_models_df <- merge(tmp2_models_df, stats_mrg_df, all.x=TRUE, sort=FALSE)
# print(tmp3_models_df)
# print(names(tmp3_models_df))
# print(glb_models_df <- subset(tmp3_models_df, select=-id.1))
# }
plt_models_df <- glb_models_df[, -grep("SD|Upper|Lower", names(glb_models_df))]
for (var in grep("^min.", names(plt_models_df), value=TRUE)) {
plt_models_df[, sub("min.", "inv.", var)] <-
#ifelse(all(is.na(tmp <- plt_models_df[, var])), NA, 1.0 / tmp)
1.0 / plt_models_df[, var]
plt_models_df <- plt_models_df[ , -grep(var, names(plt_models_df))]
}
print(plt_models_df)## id
## MFO.myMFO_classfr MFO.myMFO_classfr
## Random.myrandom_classfr Random.myrandom_classfr
## Max.cor.Y.rcv.1X1.glmnet Max.cor.Y.rcv.1X1.glmnet
## Max.cor.Y.rcv.3X1.glmnet Max.cor.Y.rcv.3X1.glmnet
## Max.cor.Y.rcv.3X3.glmnet Max.cor.Y.rcv.3X3.glmnet
## Max.cor.Y.rcv.3X5.glmnet Max.cor.Y.rcv.3X5.glmnet
## Max.cor.Y.rcv.5X1.glmnet Max.cor.Y.rcv.5X1.glmnet
## Max.cor.Y.rcv.5X3.glmnet Max.cor.Y.rcv.5X3.glmnet
## Max.cor.Y.rcv.5X5.glmnet Max.cor.Y.rcv.5X5.glmnet
## Max.cor.Y.rcv.1X1.cp.0.rpart Max.cor.Y.rcv.1X1.cp.0.rpart
## Max.cor.Y.rpart Max.cor.Y.rpart
## Interact.High.cor.Y.glmnet Interact.High.cor.Y.glmnet
## Low.cor.X.glmnet Low.cor.X.glmnet
## All.X.glmnet All.X.glmnet
## feats
## MFO.myMFO_classfr .rnorm
## Random.myrandom_classfr .rnorm
## Max.cor.Y.rcv.1X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X3.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.3X5.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X1.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X3.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.5X5.glmnet WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rcv.1X1.cp.0.rpart WordCount.root2,NDSSName.my.fctr
## Max.cor.Y.rpart WordCount.root2,NDSSName.my.fctr
## Interact.High.cor.Y.glmnet WordCount.root2,NDSSName.my.fctr,WordCount.root2:WordCount.root2
## Low.cor.X.glmnet WordCount.root2,NDSSName.my.fctr,.rnorm,WordCount.nexp
## All.X.glmnet WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns max.AUCpROC.fit max.Sens.fit
## MFO.myMFO_classfr 0 0.5000000 1.0000000
## Random.myrandom_classfr 0 0.4990604 0.8312611
## Max.cor.Y.rcv.1X1.glmnet 0 0.8790544 0.9632073
## Max.cor.Y.rcv.3X1.glmnet 25 0.8767919 0.9644760
## Max.cor.Y.rcv.3X3.glmnet 25 0.8767919 0.9644760
## Max.cor.Y.rcv.3X5.glmnet 25 0.8767919 0.9644760
## Max.cor.Y.rcv.5X1.glmnet 25 0.8784031 0.9642223
## Max.cor.Y.rcv.5X3.glmnet 25 0.8784031 0.9642223
## Max.cor.Y.rcv.5X5.glmnet 25 0.8784031 0.9642223
## Max.cor.Y.rcv.1X1.cp.0.rpart 0 0.8821543 0.9705658
## Max.cor.Y.rpart 5 0.8709432 0.9632073
## Interact.High.cor.Y.glmnet 25 0.8767919 0.9644760
## Low.cor.X.glmnet 25 0.8767919 0.9644760
## All.X.glmnet 25 0.8771175 0.9639685
## max.Spec.fit max.AUCROCR.fit
## MFO.myMFO_classfr 0.0000000 0.5000000
## Random.myrandom_classfr 0.1668598 0.4972757
## Max.cor.Y.rcv.1X1.glmnet 0.7949015 0.9608594
## Max.cor.Y.rcv.3X1.glmnet 0.7891078 0.9582555
## Max.cor.Y.rcv.3X3.glmnet 0.7891078 0.9582555
## Max.cor.Y.rcv.3X5.glmnet 0.7891078 0.9582555
## Max.cor.Y.rcv.5X1.glmnet 0.7925840 0.9607052
## Max.cor.Y.rcv.5X3.glmnet 0.7925840 0.9607052
## Max.cor.Y.rcv.5X5.glmnet 0.7925840 0.9607052
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7937428 0.9504198
## Max.cor.Y.rpart 0.7786790 0.8746354
## Interact.High.cor.Y.glmnet 0.7891078 0.9582555
## Low.cor.X.glmnet 0.7891078 0.9582555
## All.X.glmnet 0.7902665 0.9589072
## opt.prob.threshold.fit max.f.score.fit
## MFO.myMFO_classfr 0.1 0.3045703
## Random.myrandom_classfr 0.1 0.3045703
## Max.cor.Y.rcv.1X1.glmnet 0.5 0.8099174
## Max.cor.Y.rcv.3X1.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.3X3.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.3X5.glmnet 0.4 0.8099174
## Max.cor.Y.rcv.5X1.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.5X3.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.5X5.glmnet 0.5 0.8104265
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.4 0.8235294
## Max.cor.Y.rpart 0.6 0.8000000
## Interact.High.cor.Y.glmnet 0.4 0.8099174
## Low.cor.X.glmnet 0.4 0.8099174
## All.X.glmnet 0.4 0.8091873
## max.Accuracy.fit max.Kappa.fit
## MFO.myMFO_classfr 0.1796420 0.0000000
## Random.myrandom_classfr 0.1796420 0.0000000
## Max.cor.Y.rcv.1X1.glmnet 0.9329725 0.7692476
## Max.cor.Y.rcv.3X1.glmnet 0.9335973 0.7691678
## Max.cor.Y.rcv.3X3.glmnet 0.9333193 0.7690803
## Max.cor.Y.rcv.3X5.glmnet 0.9332218 0.7686375
## Max.cor.Y.rcv.5X1.glmnet 0.9331818 0.7689055
## Max.cor.Y.rcv.5X3.glmnet 0.9333905 0.7698577
## Max.cor.Y.rcv.5X5.glmnet 0.9331816 0.7691429
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9381765 0.7860827
## Max.cor.Y.rpart 0.9296422 0.7515134
## Interact.High.cor.Y.glmnet 0.9333193 0.7690803
## Low.cor.X.glmnet 0.9333193 0.7690803
## All.X.glmnet 0.9326952 0.7668828
## max.AUCpROC.OOB max.Sens.OOB max.Spec.OOB
## MFO.myMFO_classfr 0.5000000 1.0000000 0.0000000
## Random.myrandom_classfr 0.5125675 0.8077437 0.2173913
## Max.cor.Y.rcv.1X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X3.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.3X5.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X1.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X3.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.5X5.glmnet 0.5962443 0.9098798 0.2826087
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.6174697 0.9218959 0.3130435
## Max.cor.Y.rpart 0.5870523 0.9045394 0.2695652
## Interact.High.cor.Y.glmnet 0.5962443 0.9098798 0.2826087
## Low.cor.X.glmnet 0.5962443 0.9098798 0.2826087
## All.X.glmnet 0.5965780 0.9105474 0.2826087
## max.AUCROCR.OOB opt.prob.threshold.OOB
## MFO.myMFO_classfr 0.5000000 0.1
## Random.myrandom_classfr 0.4857956 0.1
## Max.cor.Y.rcv.1X1.glmnet 0.8116126 0.1
## Max.cor.Y.rcv.3X1.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.3X3.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.3X5.glmnet 0.8067975 0.1
## Max.cor.Y.rcv.5X1.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.5X3.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.5X5.glmnet 0.8114863 0.1
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7773858 0.1
## Max.cor.Y.rpart 0.5892132 0.6
## Interact.High.cor.Y.glmnet 0.8067975 0.1
## Low.cor.X.glmnet 0.8067975 0.1
## All.X.glmnet 0.8061009 0.1
## max.f.score.OOB max.Accuracy.OOB
## MFO.myMFO_classfr 0.2349336 0.1331019
## Random.myrandom_classfr 0.2349336 0.1331019
## Max.cor.Y.rcv.1X1.glmnet 0.4405405 0.7604167
## Max.cor.Y.rcv.3X1.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.3X3.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.3X5.glmnet 0.4375839 0.7575231
## Max.cor.Y.rcv.5X1.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.5X3.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.5X5.glmnet 0.4609375 0.7604167
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.4207493 0.7673611
## Max.cor.Y.rpart 0.2850575 0.8200231
## Interact.High.cor.Y.glmnet 0.4375839 0.7575231
## Low.cor.X.glmnet 0.4375839 0.7575231
## All.X.glmnet 0.4340426 0.7690972
## max.Kappa.OOB inv.elapsedtime.everything
## MFO.myMFO_classfr 0.0000000 3.7174721
## Random.myrandom_classfr 0.0000000 3.7174721
## Max.cor.Y.rcv.1X1.glmnet 0.3148374 1.0288066
## Max.cor.Y.rcv.3X1.glmnet 0.3107477 0.3538570
## Max.cor.Y.rcv.3X3.glmnet 0.3107477 0.2075981
## Max.cor.Y.rcv.3X5.glmnet 0.3107477 0.1557875
## Max.cor.Y.rcv.5X1.glmnet 0.3373693 0.3102699
## Max.cor.Y.rcv.5X3.glmnet 0.3373693 0.1628134
## Max.cor.Y.rcv.5X5.glmnet 0.3373693 0.1122586
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.2953321 1.1587486
## Max.cor.Y.rpart 0.1825002 0.3469813
## Interact.High.cor.Y.glmnet 0.3107477 0.2275313
## Low.cor.X.glmnet 0.3107477 0.2198285
## All.X.glmnet 0.3103487 0.2075981
## inv.elapsedtime.final
## MFO.myMFO_classfr 333.333333
## Random.myrandom_classfr 1000.000000
## Max.cor.Y.rcv.1X1.glmnet 3.663004
## Max.cor.Y.rcv.3X1.glmnet 3.703704
## Max.cor.Y.rcv.3X3.glmnet 3.703704
## Max.cor.Y.rcv.3X5.glmnet 3.717472
## Max.cor.Y.rcv.5X1.glmnet 3.745318
## Max.cor.Y.rcv.5X3.glmnet 3.745318
## Max.cor.Y.rcv.5X5.glmnet 3.731343
## Max.cor.Y.rcv.1X1.cp.0.rpart 14.285714
## Max.cor.Y.rpart 14.084507
## Interact.High.cor.Y.glmnet 3.703704
## Low.cor.X.glmnet 3.412969
## All.X.glmnet 3.401361print(myplot_radar(radar_inp_df=plt_models_df))## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 14. Consider specifying shapes manually if you must have them.## Warning: Removed 160 rows containing missing values (geom_point).## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 14. Consider specifying shapes manually if you must have them.
# print(myplot_radar(radar_inp_df=subset(plt_models_df,
# !(mdl_id %in% grep("random|MFO", plt_models_df$id, value=TRUE)))))
# Compute CI for <metric>SD
glb_models_df <- mutate(glb_models_df,
max.df = ifelse(max.nTuningRuns > 1, max.nTuningRuns - 1, NA),
min.sd2ci.scaler = ifelse(is.na(max.df), NA, qt(0.975, max.df)))
for (var in grep("SD", names(glb_models_df), value=TRUE)) {
# Does CI alredy exist ?
var_components <- unlist(strsplit(var, "SD"))
varActul <- paste0(var_components[1], var_components[2])
varUpper <- paste0(var_components[1], "Upper", var_components[2])
varLower <- paste0(var_components[1], "Lower", var_components[2])
if (varUpper %in% names(glb_models_df)) {
warning(varUpper, " already exists in glb_models_df")
# Assuming Lower also exists
next
}
print(sprintf("var:%s", var))
# CI is dependent on sample size in t distribution; df=n-1
glb_models_df[, varUpper] <- glb_models_df[, varActul] +
glb_models_df[, "min.sd2ci.scaler"] * glb_models_df[, var]
glb_models_df[, varLower] <- glb_models_df[, varActul] -
glb_models_df[, "min.sd2ci.scaler"] * glb_models_df[, var]
}## Warning: max.AccuracyUpper.fit already exists in glb_models_df## [1] "var:max.KappaSD.fit"# Plot metrics with CI
plt_models_df <- glb_models_df[, "id", FALSE]
pltCI_models_df <- glb_models_df[, "id", FALSE]
for (var in grep("Upper", names(glb_models_df), value=TRUE)) {
var_components <- unlist(strsplit(var, "Upper"))
col_name <- unlist(paste(var_components, collapse=""))
plt_models_df[, col_name] <- glb_models_df[, col_name]
for (name in paste0(var_components[1], c("Upper", "Lower"), var_components[2]))
pltCI_models_df[, name] <- glb_models_df[, name]
}
build_statsCI_data <- function(plt_models_df) {
mltd_models_df <- melt(plt_models_df, id.vars="id")
mltd_models_df$data <- sapply(1:nrow(mltd_models_df),
function(row_ix) tail(unlist(strsplit(as.character(
mltd_models_df[row_ix, "variable"]), "[.]")), 1))
mltd_models_df$label <- sapply(1:nrow(mltd_models_df),
function(row_ix) head(unlist(strsplit(as.character(
mltd_models_df[row_ix, "variable"]),
paste0(".", mltd_models_df[row_ix, "data"]))), 1))
#print(mltd_models_df)
return(mltd_models_df)
}
mltd_models_df <- build_statsCI_data(plt_models_df)
mltdCI_models_df <- melt(pltCI_models_df, id.vars="id")
for (row_ix in 1:nrow(mltdCI_models_df)) {
for (type in c("Upper", "Lower")) {
if (length(var_components <- unlist(strsplit(
as.character(mltdCI_models_df[row_ix, "variable"]), type))) > 1) {
#print(sprintf("row_ix:%d; type:%s; ", row_ix, type))
mltdCI_models_df[row_ix, "label"] <- var_components[1]
mltdCI_models_df[row_ix, "data"] <-
unlist(strsplit(var_components[2], "[.]"))[2]
mltdCI_models_df[row_ix, "type"] <- type
break
}
}
}
wideCI_models_df <- reshape(subset(mltdCI_models_df, select=-variable),
timevar="type",
idvar=setdiff(names(mltdCI_models_df), c("type", "value", "variable")),
direction="wide")
#print(wideCI_models_df)
mrgdCI_models_df <- merge(wideCI_models_df, mltd_models_df, all.x=TRUE)
#print(mrgdCI_models_df)
# Merge stats back in if CIs don't exist
goback_vars <- c()
for (var in unique(mltd_models_df$label)) {
for (type in unique(mltd_models_df$data)) {
var_type <- paste0(var, ".", type)
# if this data is already present, next
if (var_type %in% unique(paste(mltd_models_df$label, mltd_models_df$data,
sep=".")))
next
#print(sprintf("var_type:%s", var_type))
goback_vars <- c(goback_vars, var_type)
}
}
if (length(goback_vars) > 0) {
mltd_goback_df <- build_statsCI_data(glb_models_df[, c("id", goback_vars)])
mltd_models_df <- rbind(mltd_models_df, mltd_goback_df)
}
# mltd_models_df <- merge(mltd_models_df, glb_models_df[, c("id", "model_method")],
# all.x=TRUE)
png(paste0(glb_out_pfx, "models_bar.png"), width=480*3, height=480*2)
#print(gp <- myplot_bar(mltd_models_df, "id", "value", colorcol_name="model_method") +
print(gp <- myplot_bar(df=mltd_models_df, xcol_name="id", ycol_names="value") +
geom_errorbar(data=mrgdCI_models_df,
mapping=aes(x=mdl_id, ymax=value.Upper, ymin=value.Lower), width=0.5) +
facet_grid(label ~ data, scales="free") +
theme(axis.text.x = element_text(angle = 90,vjust = 0.5)))## Warning: Removed 4 rows containing missing values (geom_errorbar).dev.off()## quartz_off_screen
## 2print(gp)## Warning: Removed 4 rows containing missing values (geom_errorbar).
dsp_models_cols <- c("id",
glbMdlMetricsEval[glbMdlMetricsEval %in% names(glb_models_df)],
grep("opt.", names(glb_models_df), fixed = TRUE, value = TRUE))
# if (glb_is_classification && glb_is_binomial)
# dsp_models_cols <- c(dsp_models_cols, "opt.prob.threshold.OOB")
print(dsp_models_df <- orderBy(get_model_sel_frmla(), glb_models_df)[, dsp_models_cols])## id max.Accuracy.OOB
## Max.cor.Y.rpart Max.cor.Y.rpart 0.8200231
## All.X.glmnet All.X.glmnet 0.7690972
## Max.cor.Y.rcv.1X1.cp.0.rpart Max.cor.Y.rcv.1X1.cp.0.rpart 0.7673611
## Max.cor.Y.rcv.1X1.glmnet Max.cor.Y.rcv.1X1.glmnet 0.7604167
## Max.cor.Y.rcv.5X3.glmnet Max.cor.Y.rcv.5X3.glmnet 0.7604167
## Max.cor.Y.rcv.5X1.glmnet Max.cor.Y.rcv.5X1.glmnet 0.7604167
## Max.cor.Y.rcv.5X5.glmnet Max.cor.Y.rcv.5X5.glmnet 0.7604167
## Max.cor.Y.rcv.3X1.glmnet Max.cor.Y.rcv.3X1.glmnet 0.7575231
## Max.cor.Y.rcv.3X3.glmnet Max.cor.Y.rcv.3X3.glmnet 0.7575231
## Interact.High.cor.Y.glmnet Interact.High.cor.Y.glmnet 0.7575231
## Low.cor.X.glmnet Low.cor.X.glmnet 0.7575231
## Max.cor.Y.rcv.3X5.glmnet Max.cor.Y.rcv.3X5.glmnet 0.7575231
## MFO.myMFO_classfr MFO.myMFO_classfr 0.1331019
## Random.myrandom_classfr Random.myrandom_classfr 0.1331019
## max.AUCROCR.OOB max.AUCpROC.OOB
## Max.cor.Y.rpart 0.5892132 0.5870523
## All.X.glmnet 0.8061009 0.5965780
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7773858 0.6174697
## Max.cor.Y.rcv.1X1.glmnet 0.8116126 0.5962443
## Max.cor.Y.rcv.5X3.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.5X1.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.5X5.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.3X1.glmnet 0.8067975 0.5962443
## Max.cor.Y.rcv.3X3.glmnet 0.8067975 0.5962443
## Interact.High.cor.Y.glmnet 0.8067975 0.5962443
## Low.cor.X.glmnet 0.8067975 0.5962443
## Max.cor.Y.rcv.3X5.glmnet 0.8067975 0.5962443
## MFO.myMFO_classfr 0.5000000 0.5000000
## Random.myrandom_classfr 0.4857956 0.5125675
## max.Accuracy.fit opt.prob.threshold.fit
## Max.cor.Y.rpart 0.9296422 0.6
## All.X.glmnet 0.9326952 0.4
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9381765 0.4
## Max.cor.Y.rcv.1X1.glmnet 0.9329725 0.5
## Max.cor.Y.rcv.5X3.glmnet 0.9333905 0.5
## Max.cor.Y.rcv.5X1.glmnet 0.9331818 0.5
## Max.cor.Y.rcv.5X5.glmnet 0.9331816 0.5
## Max.cor.Y.rcv.3X1.glmnet 0.9335973 0.4
## Max.cor.Y.rcv.3X3.glmnet 0.9333193 0.4
## Interact.High.cor.Y.glmnet 0.9333193 0.4
## Low.cor.X.glmnet 0.9333193 0.4
## Max.cor.Y.rcv.3X5.glmnet 0.9332218 0.4
## MFO.myMFO_classfr 0.1796420 0.1
## Random.myrandom_classfr 0.1796420 0.1
## opt.prob.threshold.OOB
## Max.cor.Y.rpart 0.6
## All.X.glmnet 0.1
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.1
## Max.cor.Y.rcv.1X1.glmnet 0.1
## Max.cor.Y.rcv.5X3.glmnet 0.1
## Max.cor.Y.rcv.5X1.glmnet 0.1
## Max.cor.Y.rcv.5X5.glmnet 0.1
## Max.cor.Y.rcv.3X1.glmnet 0.1
## Max.cor.Y.rcv.3X3.glmnet 0.1
## Interact.High.cor.Y.glmnet 0.1
## Low.cor.X.glmnet 0.1
## Max.cor.Y.rcv.3X5.glmnet 0.1
## MFO.myMFO_classfr 0.1
## Random.myrandom_classfr 0.1print(myplot_radar(radar_inp_df = dsp_models_df))## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 14. Consider specifying shapes manually if you must have them.## Warning: Removed 56 rows containing missing values (geom_point).## Warning in RColorBrewer::brewer.pal(n, pal): n too large, allowed maximum for palette Set1 is 9
## Returning the palette you asked for with that many colors## Warning: The shape palette can deal with a maximum of 6 discrete values
## because more than 6 becomes difficult to discriminate; you have
## 14. Consider specifying shapes manually if you must have them.
print("Metrics used for model selection:"); print(get_model_sel_frmla())## [1] "Metrics used for model selection:"## ~-max.Accuracy.OOB - max.AUCROCR.OOB - max.AUCpROC.OOB - max.Accuracy.fit -
## opt.prob.threshold.OOB
## <environment: 0x7ff241329310>print(sprintf("Best model id: %s", dsp_models_df[1, "id"]))## [1] "Best model id: Max.cor.Y.rpart"glb_get_predictions <- function(df, mdl_id, rsp_var_out, prob_threshold_def=NULL, verbose=FALSE) {
mdl <- glb_models_lst[[mdl_id]]
#rsp_var_out <- paste0(rsp_var_out, mdl_id)
rsp_var_out <- paste0(glb_rsp_var, ".predict.")
predct_var_name <- paste0(rsp_var_out, mdl_id)
predct_prob_var_name <- paste0(rsp_var_out, mdl_id, ".prob")
predct_accurate_var_name <- paste0(rsp_var_out, mdl_id, ".accurate")
predct_error_var_name <- paste0(rsp_var_out, mdl_id, ".err")
predct_erabs_var_name <- paste0(rsp_var_out, mdl_id, ".err.abs")
if (glb_is_regression) {
df[, predct_var_name] <- predict(mdl, newdata=df, type="raw")
if (verbose) print(myplot_scatter(df, glb_rsp_var, predct_var_name) +
facet_wrap(reformulate(glb_category_var), scales = "free") +
stat_smooth(method="glm"))
df[, predct_error_var_name] <- df[, predct_var_name] - df[, glb_rsp_var]
if (verbose) print(myplot_scatter(df, predct_var_name, predct_error_var_name) +
#facet_wrap(reformulate(glb_category_var), scales = "free") +
stat_smooth(method="auto"))
if (verbose) print(myplot_scatter(df, glb_rsp_var, predct_error_var_name) +
#facet_wrap(reformulate(glb_category_var), scales = "free") +
stat_smooth(method="glm"))
df[, predct_erabs_var_name] <- abs(df[, predct_error_var_name])
if (verbose) print(head(orderBy(reformulate(c("-", predct_erabs_var_name)), df)))
df[, predct_accurate_var_name] <- (df[, glb_rsp_var] == df[, predct_var_name])
}
if (glb_is_classification && glb_is_binomial) {
prob_threshold <- glb_models_df[glb_models_df$id == mdl_id,
"opt.prob.threshold.OOB"]
if (is.null(prob_threshold) || is.na(prob_threshold)) {
warning("Using default probability threshold: ", prob_threshold_def)
if (is.null(prob_threshold <- prob_threshold_def))
stop("Default probability threshold is NULL")
}
df[, predct_prob_var_name] <- predict(mdl, newdata = df, type = "prob")[, 2]
df[, predct_var_name] <-
factor(levels(df[, glb_rsp_var])[
(df[, predct_prob_var_name] >=
prob_threshold) * 1 + 1], levels(df[, glb_rsp_var]))
# if (verbose) print(myplot_scatter(df, glb_rsp_var, predct_var_name) +
# facet_wrap(reformulate(glb_category_var), scales = "free") +
# stat_smooth(method="glm"))
df[, predct_error_var_name] <- df[, predct_var_name] != df[, glb_rsp_var]
# if (verbose) print(myplot_scatter(df, predct_var_name, predct_error_var_name) +
# #facet_wrap(reformulate(glb_category_var), scales = "free") +
# stat_smooth(method="auto"))
# if (verbose) print(myplot_scatter(df, glb_rsp_var, predct_error_var_name) +
# #facet_wrap(reformulate(glb_category_var), scales = "free") +
# stat_smooth(method="glm"))
# if prediction is a TP (true +ve), measure distance from 1.0
tp <- which((df[, predct_var_name] == df[, glb_rsp_var]) &
(df[, predct_var_name] == levels(df[, glb_rsp_var])[2]))
df[tp, predct_erabs_var_name] <- abs(1 - df[tp, predct_prob_var_name])
#rowIx <- which.max(df[tp, predct_erabs_var_name]); df[tp, c(glb_id_var, glb_rsp_var, predct_var_name, predct_prob_var_name, predct_erabs_var_name)][rowIx, ]
# if prediction is a TN (true -ve), measure distance from 0.0
tn <- which((df[, predct_var_name] == df[, glb_rsp_var]) &
(df[, predct_var_name] == levels(df[, glb_rsp_var])[1]))
df[tn, predct_erabs_var_name] <- abs(0 - df[tn, predct_prob_var_name])
#rowIx <- which.max(df[tn, predct_erabs_var_name]); df[tn, c(glb_id_var, glb_rsp_var, predct_var_name, predct_prob_var_name, predct_erabs_var_name)][rowIx, ]
# if prediction is a FP (flse +ve), measure distance from 0.0
fp <- which((df[, predct_var_name] != df[, glb_rsp_var]) &
(df[, predct_var_name] == levels(df[, glb_rsp_var])[2]))
df[fp, predct_erabs_var_name] <- abs(0 - df[fp, predct_prob_var_name])
#rowIx <- which.max(df[fp, predct_erabs_var_name]); df[fp, c(glb_id_var, glb_rsp_var, predct_var_name, predct_prob_var_name, predct_erabs_var_name)][rowIx, ]
# if prediction is a FN (flse -ve), measure distance from 1.0
fn <- which((df[, predct_var_name] != df[, glb_rsp_var]) &
(df[, predct_var_name] == levels(df[, glb_rsp_var])[1]))
df[fn, predct_erabs_var_name] <- abs(1 - df[fn, predct_prob_var_name])
#rowIx <- which.max(df[fn, predct_erabs_var_name]); df[fn, c(glb_id_var, glb_rsp_var, predct_var_name, predct_prob_var_name, predct_erabs_var_name)][rowIx, ]
if (verbose) print(head(orderBy(reformulate(c("-", predct_erabs_var_name)), df)))
df[, predct_accurate_var_name] <- (df[, glb_rsp_var] == df[, predct_var_name])
}
if (glb_is_classification && !glb_is_binomial) {
df[, predct_var_name] <- predict(mdl, newdata = df, type = "raw")
df[, paste0(predct_var_name, ".prob")] <-
predict(mdl, newdata = df, type = "prob")
stop("Multinomial prediction error calculation needs to be implemented...")
}
return(df)
}
#stop(here"); glb_to_sav(); glbObsAll <- sav_allobs_df; glbObsTrn <- sav_trnobs_df; glbObsFit <- sav_fitobs_df; glbObsOOB <- sav_OOBobs_df; sav_models_df <- glb_models_df; glb_models_df <- sav_models_df; glb_featsimp_df <- sav_featsimp_df
myget_category_stats <- function(obs_df, mdl_id, label) {
require(dplyr)
require(lazyeval)
predct_var_name <- paste0(glb_rsp_var_out, mdl_id)
predct_error_var_name <- paste0(glb_rsp_var_out, mdl_id, ".err.abs")
if (!predct_var_name %in% names(obs_df))
obs_df <- glb_get_predictions(obs_df, mdl_id, glb_rsp_var_out)
tmp_obs_df <- obs_df %>%
dplyr::select_(glb_category_var, glb_rsp_var, predct_var_name, predct_error_var_name)
#dplyr::rename(startprice.log10.predict.RFE.X.glmnet.err=error_abs_OOB)
names(tmp_obs_df)[length(names(tmp_obs_df))] <- paste0("err.abs.", label)
ret_ctgry_df <- tmp_obs_df %>%
dplyr::group_by_(glb_category_var) %>%
dplyr::summarise_(#interp(~sum(abs(var)), var=as.name(glb_rsp_var)),
interp(~sum(var), var=as.name(paste0("err.abs.", label))),
interp(~mean(var), var=as.name(paste0("err.abs.", label))),
interp(~n()))
names(ret_ctgry_df) <- c(glb_category_var,
#paste0(glb_rsp_var, ".abs.", label, ".sum"),
paste0("err.abs.", label, ".sum"),
paste0("err.abs.", label, ".mean"),
paste0(".n.", label))
ret_ctgry_df <- dplyr::ungroup(ret_ctgry_df)
#colSums(ret_ctgry_df[, -grep(glb_category_var, names(ret_ctgry_df))])
return(ret_ctgry_df)
}
#print(colSums((ctgry_df <- myget_category_stats(obs_df=glbObsFit, mdl_id="", label="fit"))[, -grep(glb_category_var, names(ctgry_df))]))
if (!is.null(glb_mdl_ensemble)) {
fit.models_2_chunk_df <- myadd_chunk(fit.models_2_chunk_df,
paste0("fit.models_2_", mdl_id_pfx), major.inc = TRUE,
label.minor = "ensemble")
mdl_id_pfx <- "Ensemble"
if (#(glb_is_regression) |
((glb_is_classification) & (!glb_is_binomial)))
stop("Ensemble models not implemented yet for multinomial classification")
mygetEnsembleAutoMdlIds <- function() {
tmp_models_df <- orderBy(get_model_sel_frmla(), glb_models_df)
row.names(tmp_models_df) <- tmp_models_df$id
mdl_threshold_pos <-
min(which(grepl("MFO|Random|Baseline", tmp_models_df$id))) - 1
mdlIds <- tmp_models_df$id[1:mdl_threshold_pos]
return(mdlIds[!grepl("Ensemble", mdlIds)])
}
if (glb_mdl_ensemble == "auto") {
glb_mdl_ensemble <- mygetEnsembleAutoMdlIds()
mdl_id_pfx <- paste0(mdl_id_pfx, ".auto")
} else if (grepl("^%<d-%", glb_mdl_ensemble)) {
glb_mdl_ensemble <- eval(parse(text =
str_trim(unlist(strsplit(glb_mdl_ensemble, "%<d-%"))[2])))
}
for (mdl_id in glb_mdl_ensemble) {
if (!(mdl_id %in% names(glb_models_lst))) {
warning("Model ", mdl_id, " in glb_model_ensemble not found !")
next
}
glbObsFit <- glb_get_predictions(df = glbObsFit, mdl_id,
glb_rsp_var_out)
glbObsOOB <- glb_get_predictions(df = glbObsOOB, mdl_id,
glb_rsp_var_out)
}
#mdl_id_pfx <- "Ensemble.RFE"; mdlId <- paste0(mdl_id_pfx, ".glmnet")
#glb_mdl_ensemble <- gsub(glb_rsp_var_out, "", grep("RFE\\.X\\.(?!Interact)", row.names(glb_featsimp_df), perl = TRUE, value = TRUE), fixed = TRUE)
#varImp(glb_models_lst[[mdlId]])
#cor_df <- data.frame(cor=cor(glbObsFit[, glb_rsp_var], glbObsFit[, paste(glb_rsp_var_out, glb_mdl_ensemble)], use="pairwise.complete.obs"))
#glbObsFit <- glb_get_predictions(df=glbObsFit, "Ensemble.glmnet", glb_rsp_var_out);print(colSums((ctgry_df <- myget_category_stats(obs_df=glbObsFit, mdl_id="Ensemble.glmnet", label="fit"))[, -grep(glb_category_var, names(ctgry_df))]))
### bid0_sp
# Better than MFO; models.n=28; min.RMSE.fit=0.0521233; err.abs.fit.sum=7.3631895
# old: Top x from auto; models.n= 5; min.RMSE.fit=0.06311047; err.abs.fit.sum=9.5937080
# RFE only ; models.n=16; min.RMSE.fit=0.05148588; err.abs.fit.sum=7.2875091
# RFE subset only ;models.n= 5; min.RMSE.fit=0.06040702; err.abs.fit.sum=9.059088
# RFE subset only ;models.n= 9; min.RMSE.fit=0.05933167; err.abs.fit.sum=8.7421288
# RFE subset only ;models.n=15; min.RMSE.fit=0.0584607; err.abs.fit.sum=8.5902066
# RFE subset only ;models.n=17; min.RMSE.fit=0.05496899; err.abs.fit.sum=8.0170431
# RFE subset only ;models.n=18; min.RMSE.fit=0.05441577; err.abs.fit.sum=7.837223
# RFE subset only ;models.n=16; min.RMSE.fit=0.05441577; err.abs.fit.sum=7.837223
### bid0_sp
### bid1_sp
# "auto"; err.abs.fit.sum=76.699774; min.RMSE.fit=0.2186429
# "RFE.X.*"; err.abs.fit.sum=; min.RMSE.fit=0.221114
### bid1_sp
indep_vars <- paste(glb_rsp_var_out, glb_mdl_ensemble, sep = "")
if (glb_is_classification)
indep_vars <- paste(indep_vars, ".prob", sep = "")
# Some models in glb_mdl_ensemble might not be fitted e.g. RFE.X.Interact
indep_vars <- intersect(indep_vars, names(glbObsFit))
# indep_vars <- grep(glb_rsp_var_out, names(glbObsFit), fixed=TRUE, value=TRUE)
# if (glb_is_regression)
# indep_vars <- indep_vars[!grepl("(err\\.abs|accurate)$", indep_vars)]
# if (glb_is_classification && glb_is_binomial)
# indep_vars <- grep("prob$", indep_vars, value=TRUE) else
# indep_vars <- indep_vars[!grepl("err$", indep_vars)]
#rfe_fit_ens_results <- myrun_rfe(glbObsFit, indep_vars)
for (method in c("glm", "glmnet")) {
for (trainControlMethod in
c("boot", "boot632", "cv", "repeatedcv"
#, "LOOCV" # tuneLength * nrow(fitDF)
, "LGOCV", "adaptive_cv"
#, "adaptive_boot" #error: adaptive$min should be less than 3
#, "adaptive_LGOCV" #error: adaptive$min should be less than 3
)) {
#sav_models_df <- glb_models_df; all.equal(sav_models_df, glb_models_df)
#glb_models_df <- sav_models_df; print(glb_models_df$id)
if ((method == "glm") && (trainControlMethod != "repeatedcv"))
# glm used only to identify outliers
next
ret_lst <- myfit_mdl(
mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = paste0(mdl_id_pfx, ".", trainControlMethod),
type = glb_model_type, tune.df = NULL,
trainControl.method = trainControlMethod,
trainControl.number = glb_rcv_n_folds,
trainControl.repeats = glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method = method)),
indep_vars = indep_vars, rsp_var = glb_rsp_var,
fit_df = glbObsFit, OOB_df = glbObsOOB)
}
}
dsp_models_df <- get_dsp_models_df()
}
if (is.null(glb_sel_mdl_id))
glb_sel_mdl_id <- dsp_models_df[1, "id"] else
print(sprintf("User specified selection: %s", glb_sel_mdl_id)) ## [1] "User specified selection: All.X.glmnet"myprint_mdl(glb_sel_mdl <- glb_models_lst[[glb_sel_mdl_id]])
## Length Class Mode
## a0 91 -none- numeric
## beta 2184 dgCMatrix S4
## df 91 -none- numeric
## dim 2 -none- numeric
## lambda 91 -none- numeric
## dev.ratio 91 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 24 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -5.52332827
## NDSSName.my.fctr#Multimedia#
## -0.16474095
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.18132553
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.90787932
## NDSSName.my.fctr#U.S.#Education
## -0.90764328
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.24502067
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.67941504
## NDSSName.my.fctrBusiness#Technology#
## 0.83419892
## NDSSName.my.fctrForeign#World#
## -0.08805297
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.83659624
## NDSSName.my.fctrOpEd#Opinion#
## 4.79265681
## NDSSName.my.fctrScience#Health#
## 3.67959505
## NDSSName.my.fctrStyles##Fashion
## -0.64461547
## NDSSName.my.fctrStyles#U.S.#
## 3.44343706
## NDSSName.my.fctrTStyle##
## -0.97236910
## NDSSName.my.fctrmyOther
## -0.23762498
## WordCount.log1p
## 0.18619880
## WordCount.root2
## 0.06691104
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -5.59881022
## NDSSName.my.fctr#Multimedia#
## -0.24609678
## NDSSName.my.fctr#Opinion#RoomForDebate
## -2.30818954
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 3.94273791
## NDSSName.my.fctr#U.S.#Education
## -1.01466873
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.30479819
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 4.69839363
## NDSSName.my.fctrBusiness#Technology#
## 0.84455935
## NDSSName.my.fctrForeign#World#
## -0.18204983
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.91223472
## NDSSName.my.fctrOpEd#Opinion#
## 4.80649138
## NDSSName.my.fctrScience#Health#
## 3.68561250
## NDSSName.my.fctrStyles##Fashion
## -0.74721366
## NDSSName.my.fctrStyles#U.S.#
## 3.44963924
## NDSSName.my.fctrTStyle##
## -1.04099014
## NDSSName.my.fctrTravel#Travel#
## -0.03203556
## NDSSName.my.fctrmyOther
## -0.35136209
## WordCount.log1p
## 0.19889748
## WordCount.root2
## 0.06700663## [1] TRUE# From here to save(), this should all be in one function
# these are executed in the same seq twice more:
# fit.data.training & predict.data.new chunks
print(sprintf("%s fit prediction diagnostics:", glb_sel_mdl_id))## [1] "All.X.glmnet fit prediction diagnostics:"glbObsFit <- glb_get_predictions(df=glbObsFit, mdl_id=glb_sel_mdl_id,
rsp_var_out=glb_rsp_var_out)
print(sprintf("%s OOB prediction diagnostics:", glb_sel_mdl_id))## [1] "All.X.glmnet OOB prediction diagnostics:"glbObsOOB <- glb_get_predictions(df = glbObsOOB, mdl_id = glb_sel_mdl_id,
rsp_var_out = glb_rsp_var_out)
glb_featsimp_df <-
myget_feats_importance(mdl=glb_sel_mdl, featsimp_df=NULL)
glb_featsimp_df[, paste0(glb_sel_mdl_id, ".importance")] <- glb_featsimp_df$importance
#mdl_id <-"RFE.X.glmnet"; glb_featsimp_df <- myget_feats_importance(glb_models_lst[[mdl_id]], glb_featsimp_df); glb_featsimp_df[, paste0(mdl_id, ".importance")] <- glb_featsimp_df$importance; print(glb_featsimp_df)
#print(head(sbst_featsimp_df <- subset(glb_featsimp_df, is.na(RFE.X.glmnet.importance) | (abs(RFE.X.YeoJohnson.glmnet.importance - RFE.X.glmnet.importance) > 0.0001), select=-importance)))
#print(orderBy(~ -cor.y.abs, subset(glb_feats_df, id %in% c(row.names(sbst_featsimp_df), "startprice.dcm1.is9", "D.weight.post.stop.sum"))))
print(glb_featsimp_df)## importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 98.38173
## NDSSName.my.fctr#Opinion#ThePublicEditor 87.34197
## NDSSName.my.fctrScience#Health# 84.05064
## NDSSName.my.fctrStyles#U.S.# 80.66804
## NDSSName.my.fctrBusiness#Technology# 43.29623
## WordCount.log1p 34.01597
## WordCount.root2 32.29795
## .rnorm 31.33944
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.33944
## NDSSName.my.fctrCulture## 31.33944
## NDSSName.my.fctrCulture#Arts# 31.33944
## NDSSName.my.fctrMetro#N.Y./Region# 31.33944
## WordCount.nexp 31.33944
## NDSSName.my.fctrTravel#Travel# 31.31570
## NDSSName.my.fctrForeign#World# 30.00852
## NDSSName.my.fctr#Multimedia# 28.91940
## NDSSName.my.fctrmyOther 27.85141
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 27.78546
## NDSSName.my.fctrStyles##Fashion 22.02991
## NDSSName.my.fctrForeign#World#AsiaPacific 19.29994
## NDSSName.my.fctr#U.S.#Education 18.25900
## NDSSName.my.fctrTStyle## 17.36032
## NDSSName.my.fctr#Opinion#RoomForDebate 0.00000
## All.X.glmnet.importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 98.38173
## NDSSName.my.fctr#Opinion#ThePublicEditor 87.34197
## NDSSName.my.fctrScience#Health# 84.05064
## NDSSName.my.fctrStyles#U.S.# 80.66804
## NDSSName.my.fctrBusiness#Technology# 43.29623
## WordCount.log1p 34.01597
## WordCount.root2 32.29795
## .rnorm 31.33944
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.33944
## NDSSName.my.fctrCulture## 31.33944
## NDSSName.my.fctrCulture#Arts# 31.33944
## NDSSName.my.fctrMetro#N.Y./Region# 31.33944
## WordCount.nexp 31.33944
## NDSSName.my.fctrTravel#Travel# 31.31570
## NDSSName.my.fctrForeign#World# 30.00852
## NDSSName.my.fctr#Multimedia# 28.91940
## NDSSName.my.fctrmyOther 27.85141
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 27.78546
## NDSSName.my.fctrStyles##Fashion 22.02991
## NDSSName.my.fctrForeign#World#AsiaPacific 19.29994
## NDSSName.my.fctr#U.S.#Education 18.25900
## NDSSName.my.fctrTStyle## 17.36032
## NDSSName.my.fctr#Opinion#RoomForDebate 0.00000# Used again in fit.data.training & predict.data.new chunks
glb_analytics_diag_plots <- function(obs_df, mdl_id, prob_threshold=NULL) {
if (!is.null(featsimp_df <- glb_featsimp_df)) {
featsimp_df$feat <- gsub("`(.*?)`", "\\1", row.names(featsimp_df))
featsimp_df$feat.interact <- gsub("(.*?):(.*)", "\\2", featsimp_df$feat)
featsimp_df$feat <- gsub("(.*?):(.*)", "\\1", featsimp_df$feat)
featsimp_df$feat.interact <-
ifelse(featsimp_df$feat.interact == featsimp_df$feat,
NA, featsimp_df$feat.interact)
featsimp_df$feat <-
gsub("(.*?)\\.fctr(.*)", "\\1\\.fctr", featsimp_df$feat)
featsimp_df$feat.interact <-
gsub("(.*?)\\.fctr(.*)", "\\1\\.fctr", featsimp_df$feat.interact)
featsimp_df <- orderBy(~ -importance.max,
summaryBy(importance ~ feat + feat.interact, data=featsimp_df,
FUN=max))
#rex_str=":(.*)"; txt_vctr=tail(featsimp_df$feat); ret_lst <- regexec(rex_str, txt_vctr); ret_lst <- regmatches(txt_vctr, ret_lst); ret_vctr <- sapply(1:length(ret_lst), function(pos_ix) ifelse(length(ret_lst[[pos_ix]]) > 0, ret_lst[[pos_ix]], "")); print(ret_vctr <- ret_vctr[ret_vctr != ""])
featsimp_df <- subset(featsimp_df, !is.na(importance.max))
if (nrow(featsimp_df) > 5) {
warning("Limiting important feature scatter plots to 5 out of ",
nrow(featsimp_df))
featsimp_df <- head(featsimp_df, 5)
}
# if (!all(is.na(featsimp_df$feat.interact)))
# stop("not implemented yet")
rsp_var_out <- paste0(glb_rsp_var_out, mdl_id)
for (var in featsimp_df$feat) {
plot_df <- melt(obs_df, id.vars = var,
measure.vars = c(glb_rsp_var, rsp_var_out))
print(myplot_scatter(plot_df, var, "value", colorcol_name = "variable",
facet_colcol_name = "variable", jitter = TRUE) +
guides(color = FALSE))
}
}
if (glb_is_regression) {
if (is.null(featsimp_df) || (nrow(featsimp_df) == 0))
warning("No important features in glb_fin_mdl") else
print(myplot_prediction_regression(df=obs_df,
feat_x=ifelse(nrow(featsimp_df) > 1, featsimp_df$feat[2],
".rownames"),
feat_y=featsimp_df$feat[1],
rsp_var=glb_rsp_var, rsp_var_out=rsp_var_out,
id_vars=glb_id_var)
# + facet_wrap(reformulate(featsimp_df$feat[2])) # if [1 or 2] is a factor
# + geom_point(aes_string(color="<col_name>.fctr")) # to color the plot
)
}
if (glb_is_classification) {
if (is.null(featsimp_df) || (nrow(featsimp_df) == 0))
warning("No features in selected model are statistically important")
else print(myplot_prediction_classification(df = obs_df,
feat_x = ifelse(nrow(featsimp_df) > 1,
featsimp_df$feat[2], ".rownames"),
feat_y = featsimp_df$feat[1],
rsp_var = glb_rsp_var,
rsp_var_out = rsp_var_out,
id_vars = glb_id_var,
prob_threshold = prob_threshold))
}
}
if (glb_is_classification && glb_is_binomial)
glb_analytics_diag_plots(obs_df = glbObsOOB, mdl_id = glb_sel_mdl_id,
prob_threshold = glb_models_df[glb_models_df$id == glb_sel_mdl_id,
"opt.prob.threshold.OOB"]) else
glb_analytics_diag_plots(obs_df=glbObsOOB, mdl_id=glb_sel_mdl_id) 
## [1] "Min/Max Boundaries: "
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 1618 1618 N 0.003961226
## 6251 6251 Y 0.933618142
## 6435 6435 N 0.110713384
## Popular.fctr.predict.All.X.glmnet
## 1618 N
## 6251 Y
## 6435 Y
## Popular.fctr.predict.All.X.glmnet.err
## 1618 FALSE
## 6251 FALSE
## 6435 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 1618 0.003961226
## 6251 0.066381858
## 6435 0.110713384
## Popular.fctr.predict.All.X.glmnet.accurate
## 1618 TRUE
## 6251 TRUE
## 6435 FALSE
## Popular.fctr.predict.All.X.glmnet.error .label
## 1618 0.00000000 1618
## 6251 0.00000000 6251
## 6435 0.01071338 6435
## [1] "Inaccurate: "
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 4020 4020 Y 0.01153216
## 4352 4352 Y 0.01557858
## 4775 4775 Y 0.01694357
## 6354 6354 Y 0.01942833
## 4745 4745 Y 0.02014592
## 172 172 Y 0.02244134
## Popular.fctr.predict.All.X.glmnet
## 4020 N
## 4352 N
## 4775 N
## 6354 N
## 4745 N
## 172 N
## Popular.fctr.predict.All.X.glmnet.err
## 4020 TRUE
## 4352 TRUE
## 4775 TRUE
## 6354 TRUE
## 4745 TRUE
## 172 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 4020 0.9884678
## 4352 0.9844214
## 4775 0.9830564
## 6354 0.9805717
## 4745 0.9798541
## 172 0.9775587
## Popular.fctr.predict.All.X.glmnet.accurate
## 4020 FALSE
## 4352 FALSE
## 4775 FALSE
## 6354 FALSE
## 4745 FALSE
## 172 FALSE
## Popular.fctr.predict.All.X.glmnet.error
## 4020 -0.08846784
## 4352 -0.08442142
## 4775 -0.08305643
## 6354 -0.08057167
## 4745 -0.07985408
## 172 -0.07755866
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 2545 2545 N 0.1014865
## 4917 4917 N 0.1072512
## 303 303 N 0.1511238
## 5087 5087 N 0.5422170
## 4962 4962 N 0.6157458
## 6511 6511 N 0.8246298
## Popular.fctr.predict.All.X.glmnet
## 2545 Y
## 4917 Y
## 303 Y
## 5087 Y
## 4962 Y
## 6511 Y
## Popular.fctr.predict.All.X.glmnet.err
## 2545 TRUE
## 4917 TRUE
## 303 TRUE
## 5087 TRUE
## 4962 TRUE
## 6511 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 2545 0.1014865
## 4917 0.1072512
## 303 0.1511238
## 5087 0.5422170
## 4962 0.6157458
## 6511 0.8246298
## Popular.fctr.predict.All.X.glmnet.accurate
## 2545 FALSE
## 4917 FALSE
## 303 FALSE
## 5087 FALSE
## 4962 FALSE
## 6511 FALSE
## Popular.fctr.predict.All.X.glmnet.error
## 2545 0.001486507
## 4917 0.007251173
## 303 0.051123801
## 5087 0.442216987
## 4962 0.515745834
## 6511 0.724629843
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 770 770 N 0.9698459
## 6276 6276 N 0.9721252
## 221 221 N 0.9729676
## 3590 3590 N 0.9730589
## 472 472 N 0.9740792
## 2995 2995 N 0.9744222
## Popular.fctr.predict.All.X.glmnet
## 770 Y
## 6276 Y
## 221 Y
## 3590 Y
## 472 Y
## 2995 Y
## Popular.fctr.predict.All.X.glmnet.err
## 770 TRUE
## 6276 TRUE
## 221 TRUE
## 3590 TRUE
## 472 TRUE
## 2995 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 770 0.9698459
## 6276 0.9721252
## 221 0.9729676
## 3590 0.9730589
## 472 0.9740792
## 2995 0.9744222
## Popular.fctr.predict.All.X.glmnet.accurate
## 770 FALSE
## 6276 FALSE
## 221 FALSE
## 3590 FALSE
## 472 FALSE
## 2995 FALSE
## Popular.fctr.predict.All.X.glmnet.error
## 770 0.8698459
## 6276 0.8721252
## 221 0.8729676
## 3590 0.8730589
## 472 0.8740792
## 2995 0.8744222
if (!is.null(glb_category_var)) {
glb_ctgry_df <- merge(glb_ctgry_df,
myget_category_stats(obs_df = glbObsFit, mdl_id = glb_sel_mdl_id, label = "fit"),
by = glb_category_var, all = TRUE)
row.names(glb_ctgry_df) <- glb_ctgry_df[, glb_category_var]
glb_ctgry_df <- merge(glb_ctgry_df,
myget_category_stats(obs_df=glbObsOOB, mdl_id=glb_sel_mdl_id, label="OOB"),
#by=glb_category_var, all=TRUE) glb_ctgry-df already contains .n.OOB ?
all=TRUE)
row.names(glb_ctgry_df) <- glb_ctgry_df[, glb_category_var]
if (any(grepl("OOB", glbMdlMetricsEval)))
print(orderBy(~-err.abs.OOB.mean, glb_ctgry_df)) else
print(orderBy(~-err.abs.fit.mean, glb_ctgry_df))
print(colSums(glb_ctgry_df[, -grep(glb_category_var, names(glb_ctgry_df))]))
}## NDSSName.my.fctr
## OpEd#Opinion# OpEd#Opinion#
## Styles#U.S.# Styles#U.S.#
## Science#Health# Science#Health#
## Business#Crosswords/Games# Business#Crosswords/Games#
## #Opinion#ThePublicEditor #Opinion#ThePublicEditor
## Business#Technology# Business#Technology#
## Business#BusinessDay#Dealbook Business#BusinessDay#Dealbook
## ## ##
## Metro#N.Y./Region# Metro#N.Y./Region#
## Culture#Arts# Culture#Arts#
## Business#BusinessDay#SmallBusiness Business#BusinessDay#SmallBusiness
## Styles##Fashion Styles##Fashion
## #Opinion#RoomForDebate #Opinion#RoomForDebate
## #Multimedia# #Multimedia#
## Travel#Travel# Travel#Travel#
## TStyle## TStyle##
## Foreign#World#AsiaPacific Foreign#World#AsiaPacific
## myOther myOther
## Culture## Culture##
## #U.S.#Education #U.S.#Education
## Foreign#World# Foreign#World#
## .n.OOB .n.Fit .n.Tst .freqRatio.Fit
## OpEd#Opinion# 89 437 164 0.090965862
## Styles#U.S.# 50 127 61 0.026436303
## Science#Health# 48 148 57 0.030807660
## Business#Crosswords/Games# 18 105 42 0.021856786
## #Opinion#ThePublicEditor 4 16 10 0.003330558
## Business#Technology# 126 213 114 0.044338052
## Business#BusinessDay#Dealbook 323 629 304 0.130932556
## ## 371 913 342 0.190049958
## Metro#N.Y./Region# 70 128 67 0.026644463
## Culture#Arts# 185 490 174 0.101998335
## Business#BusinessDay#SmallBusiness 40 100 41 0.020815987
## Styles##Fashion 15 104 15 0.021648626
## #Opinion#RoomForDebate 20 42 20 0.008742714
## #Multimedia# 49 92 52 0.019150708
## Travel#Travel# 34 83 35 0.017277269
## TStyle## 101 623 105 0.129683597
## Foreign#World#AsiaPacific 53 150 56 0.031223980
## myOther 5 33 5 0.006869276
## Culture## 1 NA 70 NA
## #U.S.#Education 82 243 89 0.050582848
## Foreign#World# 44 128 47 0.026644463
## .freqRatio.OOB .freqRatio.Tst
## OpEd#Opinion# 0.0515046296 0.087700535
## Styles#U.S.# 0.0289351852 0.032620321
## Science#Health# 0.0277777778 0.030481283
## Business#Crosswords/Games# 0.0104166667 0.022459893
## #Opinion#ThePublicEditor 0.0023148148 0.005347594
## Business#Technology# 0.0729166667 0.060962567
## Business#BusinessDay#Dealbook 0.1869212963 0.162566845
## ## 0.2146990741 0.182887701
## Metro#N.Y./Region# 0.0405092593 0.035828877
## Culture#Arts# 0.1070601852 0.093048128
## Business#BusinessDay#SmallBusiness 0.0231481481 0.021925134
## Styles##Fashion 0.0086805556 0.008021390
## #Opinion#RoomForDebate 0.0115740741 0.010695187
## #Multimedia# 0.0283564815 0.027807487
## Travel#Travel# 0.0196759259 0.018716578
## TStyle## 0.0584490741 0.056149733
## Foreign#World#AsiaPacific 0.0306712963 0.029946524
## myOther 0.0028935185 0.002673797
## Culture## 0.0005787037 0.037433155
## #U.S.#Education 0.0474537037 0.047593583
## Foreign#World# 0.0254629630 0.025133690
## err.abs.fit.sum err.abs.fit.mean .n.fit
## OpEd#Opinion# 101.962286 0.23332331 437
## Styles#U.S.# 51.707342 0.40714443 127
## Science#Health# 48.666469 0.32882749 148
## Business#Crosswords/Games# 19.346059 0.18424818 105
## #Opinion#ThePublicEditor 4.044658 0.25279116 16
## Business#Technology# 46.446343 0.21805795 213
## Business#BusinessDay#Dealbook 76.983498 0.12239030 629
## ## 91.828632 0.10057901 913
## Metro#N.Y./Region# 14.444993 0.11285150 128
## Culture#Arts# 41.599488 0.08489691 490
## Business#BusinessDay#SmallBusiness 8.607239 0.08607239 100
## Styles##Fashion 3.033172 0.02916512 104
## #Opinion#RoomForDebate 1.938007 0.04614303 42
## #Multimedia# 4.739265 0.05151375 92
## Travel#Travel# 2.569132 0.03095340 83
## TStyle## 14.962143 0.02401628 623
## Foreign#World#AsiaPacific 7.489991 0.04993328 150
## myOther 1.720179 0.05212665 33
## Culture## NA NA NA
## #U.S.#Education 4.569704 0.01880537 243
## Foreign#World# 3.357443 0.02623002 128
## err.abs.OOB.sum err.abs.OOB.mean
## OpEd#Opinion# 52.11236621 0.58553220
## Styles#U.S.# 27.39861838 0.54797237
## Science#Health# 25.12604617 0.52345930
## Business#Crosswords/Games# 9.01410297 0.50078350
## #Opinion#ThePublicEditor 1.97530353 0.49382588
## Business#Technology# 28.55232120 0.22660572
## Business#BusinessDay#Dealbook 57.75141149 0.17879694
## ## 63.39672286 0.17088065
## Metro#N.Y./Region# 11.12676976 0.15895385
## Culture#Arts# 28.98681899 0.15668551
## Business#BusinessDay#SmallBusiness 3.84496527 0.09612413
## Styles##Fashion 1.42711318 0.09514088
## #Opinion#RoomForDebate 1.78910542 0.08945527
## #Multimedia# 2.85858104 0.05833839
## Travel#Travel# 1.97769164 0.05816740
## TStyle## 5.15834308 0.05107270
## Foreign#World#AsiaPacific 2.69695832 0.05088601
## myOther 0.23180569 0.04636114
## Culture## 0.02849447 0.02849447
## #U.S.#Education 2.22191005 0.02709646
## Foreign#World# 1.13916144 0.02589003
## .n.OOB .n.Fit .n.Tst .freqRatio.Fit
## 1728.000000 NA 1870.000000 NA
## .freqRatio.OOB .freqRatio.Tst err.abs.fit.sum err.abs.fit.mean
## 1.000000 1.000000 NA NA
## .n.fit err.abs.OOB.sum err.abs.OOB.mean
## NA 328.814611 4.170523write.csv(glbObsOOB[, c(glb_id_var,
grep(glb_rsp_var, names(glbObsOOB), fixed=TRUE, value=TRUE))],
paste0(gsub(".", "_", paste0(glb_out_pfx, glb_sel_mdl_id), fixed=TRUE),
"_OOBobs.csv"), row.names=FALSE)
fit.models_2_chunk_df <-
myadd_chunk(NULL, "fit.models_2_bgn", label.minor = "teardown")## label step_major step_minor label_minor bgn end elapsed
## 1 fit.models_2_bgn 1 0 teardown 174.883 NA NAglb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.models", major.inc=FALSE)## label step_major step_minor label_minor bgn end elapsed
## 12 fit.models 6 2 2 161.330 174.893 13.563
## 13 fit.models 6 3 3 174.894 NA NA# if (sum(is.na(glbObsAll$D.P.http)) > 0)
# stop("fit.models_3: Why is this happening ?")
#stop(here"); glb_to_sav()
sync_glb_obs_df <- function() {
# Merge or cbind ?
for (col in setdiff(names(glbObsFit), names(glbObsTrn)))
glbObsTrn[glbObsTrn$.lcn == "Fit", col] <<- glbObsFit[, col]
for (col in setdiff(names(glbObsFit), names(glbObsAll)))
glbObsAll[glbObsAll$.lcn == "Fit", col] <<- glbObsFit[, col]
if (all(is.na(glbObsNew[, glb_rsp_var])))
for (col in setdiff(names(glbObsOOB), names(glbObsTrn)))
glbObsTrn[glbObsTrn$.lcn == "OOB", col] <<- glbObsOOB[, col]
for (col in setdiff(names(glbObsOOB), names(glbObsAll)))
glbObsAll[glbObsAll$.lcn == "OOB", col] <<- glbObsOOB[, col]
}
sync_glb_obs_df()
print(setdiff(names(glbObsNew), names(glbObsAll)))## character(0)if (glb_save_envir)
save(glb_feats_df,
glbObsAll, #glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
glb_models_df, dsp_models_df, glb_models_lst, glb_sel_mdl, glb_sel_mdl_id,
glb_model_type,
file=paste0(glb_out_pfx, "selmdl_dsk.RData"))
#load(paste0(glb_out_pfx, "selmdl_dsk.RData"))
rm(ret_lst)## Warning in rm(ret_lst): object 'ret_lst' not foundreplay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"model.selected")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
## 2.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction data.new.prediction firing: model.selected
## 3.0000 3 0 2 1 0
glb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.data.training", major.inc=TRUE)## label step_major step_minor label_minor bgn end
## 13 fit.models 6 3 3 174.894 181.343
## 14 fit.data.training 7 0 0 181.344 NA
## elapsed
## 13 6.45
## 14 NAStep 7.0: fit data training
#load(paste0(glb_inp_pfx, "dsk.RData"))
if (!is.null(glb_fin_mdl_id) && (glb_fin_mdl_id %in% names(glb_models_lst))) {
warning("Final model same as user selected model")
glb_fin_mdl <- glb_models_lst[[glb_fin_mdl_id]]
} else
# if (nrow(glbObsFit) + length(glbObsFitOutliers) == nrow(glbObsTrn))
if (!all(is.na(glbObsNew[, glb_rsp_var])))
{
warning("Final model same as glb_sel_mdl_id")
glb_fin_mdl_id <- paste0("Final.", glb_sel_mdl_id)
glb_fin_mdl <- glb_sel_mdl
glb_models_lst[[glb_fin_mdl_id]] <- glb_fin_mdl
} else {
# if (grepl("RFE", glb_sel_mdl_id) ||
# (!is.null(glb_mdl_ensemble) && grepl("RFE", glb_mdl_ensemble))) {
indep_vars <- myadjust_interaction_feats(subset(glb_feats_df,
!nzv & (exclude.as.feat != 1))[, "id"])
rfe_trn_results <-
myrun_rfe(glbObsTrn, indep_vars, glbRFESizes[["Final"]])
if (!isTRUE(all.equal(sort(predictors(rfe_trn_results)),
sort(predictors(rfe_fit_results))))) {
print("Diffs predictors(rfe_trn_results) vs. predictors(rfe_fit_results):")
print(setdiff(predictors(rfe_trn_results), predictors(rfe_fit_results)))
print("Diffs predictors(rfe_fit_results) vs. predictors(rfe_trn_results):")
print(setdiff(predictors(rfe_fit_results), predictors(rfe_trn_results)))
}
# }
if (grepl("Ensemble", glb_sel_mdl_id)) {
# Find which models are relevant
mdlimp_df <- subset(myget_feats_importance(glb_sel_mdl), importance > 5)
# Fit selected models on glbObsTrn
for (mdl_id in gsub(".prob", "",
gsub(glb_rsp_var_out, "", row.names(mdlimp_df), fixed = TRUE),
fixed = TRUE)) {
mdl_id_components <- unlist(strsplit(mdl_id, "[.]"))
mdlIdPfx <- paste0(c(head(mdl_id_components, -1), "Train"),
collapse = ".")
if (grepl("RFE\\.X\\.", mdlIdPfx))
mdlIndepVars <- myadjust_interaction_feats(myextract_actual_feats(
predictors(rfe_trn_results))) else
mdlIndepVars <- trim(unlist(
strsplit(glb_models_df[glb_models_df$id == mdl_id, "feats"], "[,]")))
ret_lst <-
myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = mdlIdPfx,
type = glb_model_type, tune.df = glb_tune_models_df,
trainControl.method = "repeatedcv",
trainControl.number = glb_rcv_n_folds,
trainControl.repeats = glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method = tail(mdl_id_components, 1))),
indep_vars = mdlIndepVars,
rsp_var = glb_rsp_var,
fit_df = glbObsTrn, OOB_df = NULL)
glbObsTrn <- glb_get_predictions(df = glbObsTrn,
mdl_id = tail(glb_models_df$id, 1),
rsp_var_out = glb_rsp_var_out,
prob_threshold_def =
subset(glb_models_df, id == mdl_id)$opt.prob.threshold.OOB)
glbObsNew <- glb_get_predictions(df = glbObsNew,
mdl_id = tail(glb_models_df$id, 1),
rsp_var_out = glb_rsp_var_out,
prob_threshold_def =
subset(glb_models_df, id == mdl_id)$opt.prob.threshold.OOB)
}
}
# "Final" model
if ((model_method <- glb_sel_mdl$method) == "custom")
# get actual method from the mdl_id
model_method <- tail(unlist(strsplit(glb_sel_mdl_id, "[.]")), 1)
if (grepl("Ensemble", glb_sel_mdl_id)) {
# Find which models are relevant
mdlimp_df <- subset(myget_feats_importance(glb_sel_mdl), importance > 5)
if (glb_is_classification && glb_is_binomial)
indep_vars_vctr <- gsub("(.*)\\.(.*)\\.prob", "\\1\\.Train\\.\\2\\.prob",
row.names(mdlimp_df)) else
indep_vars_vctr <- gsub("(.*)\\.(.*)", "\\1\\.Train\\.\\2",
row.names(mdlimp_df))
} else
if (grepl("RFE.X", glb_sel_mdl_id, fixed = TRUE)) {
indep_vars_vctr <- myextract_actual_feats(predictors(rfe_trn_results))
} else indep_vars_vctr <-
trim(unlist(strsplit(glb_models_df[glb_models_df$id ==
glb_sel_mdl_id
, "feats"], "[,]")))
# Discontinuing use of tune_finmdl_df;
# since final model needs to be cved on glbObsTrn
# tune_finmdl_df <- NULL
# if (nrow(glb_sel_mdl$bestTune) > 0) {
# for (param in names(glb_sel_mdl$bestTune)) {
# #print(sprintf("param: %s", param))
# if (glb_sel_mdl$bestTune[1, param] != "none")
# tune_finmdl_df <- rbind(tune_finmdl_df,
# data.frame(parameter=param,
# min=glb_sel_mdl$bestTune[1, param],
# max=glb_sel_mdl$bestTune[1, param],
# by=1)) # by val does not matter
# }
# }
# Sync with parameters in mydsutils.R
#stop(here"); glb_to_sav(); glb_models_lst <- sav_models_lst; glb_models_df <- sav_models_df
if (!is.null(glb_preproc_methods) &&
((match_pos <- regexpr(gsub(".", "\\.",
paste(glb_preproc_methods, collapse = "|"),
fixed = TRUE), glb_sel_mdl_id)) != -1))
ths_preProcess <- str_sub(glb_sel_mdl_id, match_pos,
match_pos + attr(match_pos, "match.length") - 1) else
ths_preProcess <- NULL
mdl_id_pfx <- ifelse(grepl("Ensemble", glb_sel_mdl_id),
"Final.Ensemble", "Final")
trnobs_df <- if (is.null(glbObsTrnOutliers[[mdl_id_pfx]])) glbObsTrn else
glbObsTrn[!(glbObsTrn[, glb_id_var] %in%
glbObsTrnOutliers[[mdl_id_pfx]]), ]
# Force fitting of Final.glm to identify outliers
method_vctr <-
unique(c("glm", tail(unlist(strsplit(glb_sel_mdl_id, "[.]")), 1)))
for (method in method_vctr) {
#source("caret_nominalTrainWorkflow.R")
# glmnet requires at least 2 indep vars
if ((length(indep_vars_vctr) == 1) && (method %in% "glmnet"))
next
ret_lst <-
myfit_mdl(mdl_specs_lst = myinit_mdl_specs_lst(mdl_specs_lst = list(
id.prefix = mdl_id_pfx,
type = glb_model_type, trainControl.method = "repeatedcv",
trainControl.number = glb_rcv_n_folds,
trainControl.repeats = glb_rcv_n_repeats,
trainControl.classProbs = glb_is_classification,
trainControl.summaryFunction = glbMdlMetricSummaryFn,
trainControl.allowParallel = FALSE,
train.metric = glbMdlMetricSummary,
train.maximize = glbMdlMetricMaximize,
train.method = method,
train.preProcess = ths_preProcess)),
indep_vars = indep_vars_vctr, rsp_var = glb_rsp_var,
fit_df = trnobs_df, OOB_df = NULL)
}
if ((length(method_vctr) == 1) || (method != "glm")) {
glb_fin_mdl <- glb_models_lst[[length(glb_models_lst)]]
glb_fin_mdl_id <- glb_models_df[length(glb_models_lst), "id"]
}
}## +(rfe) fit Fold1.Rep1 size: 23
## -(rfe) fit Fold1.Rep1 size: 23
## +(rfe) imp Fold1.Rep1
## -(rfe) imp Fold1.Rep1
## +(rfe) fit Fold1.Rep1 size: 16
## -(rfe) fit Fold1.Rep1 size: 16
## +(rfe) fit Fold1.Rep1 size: 8
## -(rfe) fit Fold1.Rep1 size: 8
## +(rfe) fit Fold1.Rep1 size: 4
## -(rfe) fit Fold1.Rep1 size: 4
## +(rfe) fit Fold1.Rep1 size: 2
## -(rfe) fit Fold1.Rep1 size: 2
## +(rfe) fit Fold2.Rep1 size: 23
## -(rfe) fit Fold2.Rep1 size: 23
## +(rfe) imp Fold2.Rep1
## -(rfe) imp Fold2.Rep1
## +(rfe) fit Fold2.Rep1 size: 16
## -(rfe) fit Fold2.Rep1 size: 16
## +(rfe) fit Fold2.Rep1 size: 8
## -(rfe) fit Fold2.Rep1 size: 8
## +(rfe) fit Fold2.Rep1 size: 4
## -(rfe) fit Fold2.Rep1 size: 4
## +(rfe) fit Fold2.Rep1 size: 2
## -(rfe) fit Fold2.Rep1 size: 2
## +(rfe) fit Fold3.Rep1 size: 23
## -(rfe) fit Fold3.Rep1 size: 23
## +(rfe) imp Fold3.Rep1
## -(rfe) imp Fold3.Rep1
## +(rfe) fit Fold3.Rep1 size: 16
## -(rfe) fit Fold3.Rep1 size: 16
## +(rfe) fit Fold3.Rep1 size: 8
## -(rfe) fit Fold3.Rep1 size: 8
## +(rfe) fit Fold3.Rep1 size: 4
## -(rfe) fit Fold3.Rep1 size: 4
## +(rfe) fit Fold3.Rep1 size: 2
## -(rfe) fit Fold3.Rep1 size: 2
## +(rfe) fit Fold1.Rep2 size: 23
## -(rfe) fit Fold1.Rep2 size: 23
## +(rfe) imp Fold1.Rep2
## -(rfe) imp Fold1.Rep2
## +(rfe) fit Fold1.Rep2 size: 16
## -(rfe) fit Fold1.Rep2 size: 16
## +(rfe) fit Fold1.Rep2 size: 8
## -(rfe) fit Fold1.Rep2 size: 8
## +(rfe) fit Fold1.Rep2 size: 4
## -(rfe) fit Fold1.Rep2 size: 4
## +(rfe) fit Fold1.Rep2 size: 2
## -(rfe) fit Fold1.Rep2 size: 2
## +(rfe) fit Fold2.Rep2 size: 23
## -(rfe) fit Fold2.Rep2 size: 23
## +(rfe) imp Fold2.Rep2
## -(rfe) imp Fold2.Rep2
## +(rfe) fit Fold2.Rep2 size: 16
## -(rfe) fit Fold2.Rep2 size: 16
## +(rfe) fit Fold2.Rep2 size: 8
## -(rfe) fit Fold2.Rep2 size: 8
## +(rfe) fit Fold2.Rep2 size: 4
## -(rfe) fit Fold2.Rep2 size: 4
## +(rfe) fit Fold2.Rep2 size: 2
## -(rfe) fit Fold2.Rep2 size: 2
## +(rfe) fit Fold3.Rep2 size: 23
## -(rfe) fit Fold3.Rep2 size: 23
## +(rfe) imp Fold3.Rep2
## -(rfe) imp Fold3.Rep2
## +(rfe) fit Fold3.Rep2 size: 16
## -(rfe) fit Fold3.Rep2 size: 16
## +(rfe) fit Fold3.Rep2 size: 8
## -(rfe) fit Fold3.Rep2 size: 8
## +(rfe) fit Fold3.Rep2 size: 4
## -(rfe) fit Fold3.Rep2 size: 4
## +(rfe) fit Fold3.Rep2 size: 2
## -(rfe) fit Fold3.Rep2 size: 2
## +(rfe) fit Fold1.Rep3 size: 23
## -(rfe) fit Fold1.Rep3 size: 23
## +(rfe) imp Fold1.Rep3
## -(rfe) imp Fold1.Rep3
## +(rfe) fit Fold1.Rep3 size: 16
## -(rfe) fit Fold1.Rep3 size: 16
## +(rfe) fit Fold1.Rep3 size: 8
## -(rfe) fit Fold1.Rep3 size: 8
## +(rfe) fit Fold1.Rep3 size: 4
## -(rfe) fit Fold1.Rep3 size: 4
## +(rfe) fit Fold1.Rep3 size: 2
## -(rfe) fit Fold1.Rep3 size: 2
## +(rfe) fit Fold2.Rep3 size: 23
## -(rfe) fit Fold2.Rep3 size: 23
## +(rfe) imp Fold2.Rep3
## -(rfe) imp Fold2.Rep3
## +(rfe) fit Fold2.Rep3 size: 16
## -(rfe) fit Fold2.Rep3 size: 16
## +(rfe) fit Fold2.Rep3 size: 8
## -(rfe) fit Fold2.Rep3 size: 8
## +(rfe) fit Fold2.Rep3 size: 4
## -(rfe) fit Fold2.Rep3 size: 4
## +(rfe) fit Fold2.Rep3 size: 2
## -(rfe) fit Fold2.Rep3 size: 2
## +(rfe) fit Fold3.Rep3 size: 23
## -(rfe) fit Fold3.Rep3 size: 23
## +(rfe) imp Fold3.Rep3
## -(rfe) imp Fold3.Rep3
## +(rfe) fit Fold3.Rep3 size: 16
## -(rfe) fit Fold3.Rep3 size: 16
## +(rfe) fit Fold3.Rep3 size: 8
## -(rfe) fit Fold3.Rep3 size: 8
## +(rfe) fit Fold3.Rep3 size: 4
## -(rfe) fit Fold3.Rep3 size: 4
## +(rfe) fit Fold3.Rep3 size: 2
## -(rfe) fit Fold3.Rep3 size: 2
##
## Recursive feature selection
##
## Outer resampling method: Cross-Validated (3 fold, repeated 3 times)
##
## Resampling performance over subset size:
##
## Variables Accuracy Kappa AccuracySD KappaSD Selected
## 2 0.8204 0.03718 0.003445 0.01035
## 4 0.8760 0.44502 0.003122 0.01919
## 8 0.9018 0.63845 0.006472 0.02305
## 16 0.9016 0.63798 0.006364 0.02240
## 23 0.9033 0.64712 0.006639 0.02346 *
##
## The top 5 variables (out of 23):
## WordCount.log1p, WordCount.root2, WordCount.nexp, NDSSName.my.fctrOpEd#Opinion#, NDSSName.my.fctrScience#Health#
##
## [1] "WordCount.log1p"
## [2] "WordCount.root2"
## [3] "WordCount.nexp"
## [4] "NDSSName.my.fctrOpEd#Opinion#"
## [5] "NDSSName.my.fctrScience#Health#"
## [6] "NDSSName.my.fctrBusiness#Crosswords/Games#"
## [7] "NDSSName.my.fctrStyles#U.S.#"
## [8] ".rnorm"
## [9] "NDSSName.my.fctrBusiness#Technology#"
## [10] "NDSSName.my.fctrmyOther"
## [11] "NDSSName.my.fctr#Opinion#RoomForDebate"
## [12] "NDSSName.my.fctrMetro#N.Y./Region#"
## [13] "NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness"
## [14] "NDSSName.my.fctrTravel#Travel#"
## [15] "NDSSName.my.fctrStyles##Fashion"
## [16] "NDSSName.my.fctr#Multimedia#"
## [17] "NDSSName.my.fctrForeign#World#"
## [18] "NDSSName.my.fctrForeign#World#AsiaPacific"
## [19] "NDSSName.my.fctr#U.S.#Education"
## [20] "NDSSName.my.fctrCulture#Arts#"
## [21] "NDSSName.my.fctrBusiness#BusinessDay#Dealbook"
## [22] "NDSSName.my.fctr##"
## [23] "NDSSName.my.fctrTStyle##"## [1] "fitting model: Final.glm"
## [1] " indep_vars: WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp"
## + Fold1.Rep1: parameter=none## Warning: glm.fit: algorithm did not converge## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## - Fold1.Rep1: parameter=none
## + Fold2.Rep1: parameter=none## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold2.Rep1: parameter=none
## + Fold3.Rep1: parameter=none
## - Fold3.Rep1: parameter=none
## + Fold1.Rep2: parameter=none
## - Fold1.Rep2: parameter=none
## + Fold2.Rep2: parameter=none
## - Fold2.Rep2: parameter=none
## + Fold3.Rep2: parameter=none## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold3.Rep2: parameter=none
## + Fold1.Rep3: parameter=none## Warning in predict.lm(object, newdata, se.fit, scale = 1, type =
## ifelse(type == : prediction from a rank-deficient fit may be misleading## - Fold1.Rep3: parameter=none
## + Fold2.Rep3: parameter=none
## - Fold2.Rep3: parameter=none
## + Fold3.Rep3: parameter=none## Warning: glm.fit: algorithm did not converge## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred## - Fold3.Rep3: parameter=none
## Aggregating results
## Fitting final model on full training set## Warning: not plotting observations with leverage one:
## 4746
## Warning: not plotting observations with leverage one:
## 4746
##
## Call:
## NULL
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -2.6763 -0.3646 -0.2216 -0.0040 3.4318
##
## Coefficients:
## Estimate Std. Error
## (Intercept) -6.375e+00 7.803e-01
## .rnorm -8.854e-03 4.645e-02
## `NDSSName.my.fctr#Multimedia#` -1.972e+00 7.249e-01
## `NDSSName.my.fctr#Opinion#RoomForDebate` -4.099e+00 1.029e+00
## `NDSSName.my.fctr#Opinion#ThePublicEditor` 3.212e+00 5.811e-01
## `NDSSName.my.fctr#U.S.#Education` -1.623e+01 3.335e+02
## `NDSSName.my.fctrBusiness#BusinessDay#Dealbook` -4.406e-01 1.572e-01
## `NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness` -1.444e+00 4.697e-01
## `NDSSName.my.fctrBusiness#Crosswords/Games#` 3.786e+00 2.732e-01
## `NDSSName.my.fctrBusiness#Technology#` 4.470e-01 1.896e-01
## `NDSSName.my.fctrCulture##` -1.549e+01 6.523e+03
## `NDSSName.my.fctrCulture#Arts#` -5.450e-02 1.876e-01
## `NDSSName.my.fctrForeign#World#` -1.550e+01 4.776e+02
## `NDSSName.my.fctrForeign#World#AsiaPacific` -2.497e+00 5.941e-01
## `NDSSName.my.fctrMetro#N.Y./Region#` -1.490e-01 2.852e-01
## `NDSSName.my.fctrOpEd#Opinion#` 3.742e+00 1.616e-01
## `NDSSName.my.fctrScience#Health#` 2.670e+00 1.905e-01
## `NDSSName.my.fctrStyles##Fashion` -2.572e+00 1.011e+00
## `NDSSName.my.fctrStyles#U.S.#` 2.358e+00 1.952e-01
## `NDSSName.my.fctrTStyle##` -2.135e+00 3.738e-01
## `NDSSName.my.fctrTravel#Travel#` -1.816e+00 1.012e+00
## NDSSName.my.fctrmyOther -1.650e+01 9.845e+02
## WordCount.log1p 4.480e-01 1.827e-01
## WordCount.nexp -4.904e+00 2.796e+01
## WordCount.root2 6.029e-02 1.631e-02
## z value Pr(>|z|)
## (Intercept) -8.170 3.09e-16 ***
## .rnorm -0.191 0.848831
## `NDSSName.my.fctr#Multimedia#` -2.720 0.006521 **
## `NDSSName.my.fctr#Opinion#RoomForDebate` -3.985 6.76e-05 ***
## `NDSSName.my.fctr#Opinion#ThePublicEditor` 5.528 3.24e-08 ***
## `NDSSName.my.fctr#U.S.#Education` -0.049 0.961192
## `NDSSName.my.fctrBusiness#BusinessDay#Dealbook` -2.803 0.005068 **
## `NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness` -3.075 0.002103 **
## `NDSSName.my.fctrBusiness#Crosswords/Games#` 13.862 < 2e-16 ***
## `NDSSName.my.fctrBusiness#Technology#` 2.358 0.018384 *
## `NDSSName.my.fctrCulture##` -0.002 0.998106
## `NDSSName.my.fctrCulture#Arts#` -0.290 0.771462
## `NDSSName.my.fctrForeign#World#` -0.032 0.974115
## `NDSSName.my.fctrForeign#World#AsiaPacific` -4.204 2.63e-05 ***
## `NDSSName.my.fctrMetro#N.Y./Region#` -0.522 0.601428
## `NDSSName.my.fctrOpEd#Opinion#` 23.151 < 2e-16 ***
## `NDSSName.my.fctrScience#Health#` 14.015 < 2e-16 ***
## `NDSSName.my.fctrStyles##Fashion` -2.544 0.010969 *
## `NDSSName.my.fctrStyles#U.S.#` 12.084 < 2e-16 ***
## `NDSSName.my.fctrTStyle##` -5.710 1.13e-08 ***
## `NDSSName.my.fctrTravel#Travel#` -1.795 0.072576 .
## NDSSName.my.fctrmyOther -0.017 0.986627
## WordCount.log1p 2.452 0.014212 *
## WordCount.nexp -0.175 0.860772
## WordCount.root2 3.698 0.000218 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for binomial family taken to be 1)
##
## Null deviance: 5900.1 on 6531 degrees of freedom
## Residual deviance: 3110.5 on 6507 degrees of freedom
## AIC: 3160.5
##
## Number of Fisher Scoring iterations: 17
## Popular.fctr Popular.fctr.predict.Final.glm.N
## 1 N 5172
## 2 Y 363
## Popular.fctr.predict.Final.glm.Y
## 1 267
## 2 730
## Prediction
## Reference N Y
## N 5172 267
## Y 363 730
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.035517e-01 6.413003e-01 8.961350e-01 9.106059e-01 8.326699e-01
## AccuracyPValue McnemarPValue
## 7.824873e-61 1.537762e-04
## id
## 1 Final.glm
## feats
## 1 WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 1 3.307 0.261
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.7989009 0.9610222 0.6367795 0.9333994
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.4 0.6985646 0.9057971
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.896135 0.9106059 0.6371833
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.004986598 0.02341093
## [1] "fitting model: Final.glmnet"
## [1] " indep_vars: WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp"
## + Fold1.Rep1: alpha=0.100, lambda=0.07781
## - Fold1.Rep1: alpha=0.100, lambda=0.07781
## + Fold1.Rep1: alpha=0.325, lambda=0.07781
## - Fold1.Rep1: alpha=0.325, lambda=0.07781
## + Fold1.Rep1: alpha=0.550, lambda=0.07781
## - Fold1.Rep1: alpha=0.550, lambda=0.07781
## + Fold1.Rep1: alpha=0.775, lambda=0.07781
## - Fold1.Rep1: alpha=0.775, lambda=0.07781
## + Fold1.Rep1: alpha=1.000, lambda=0.07781
## - Fold1.Rep1: alpha=1.000, lambda=0.07781
## + Fold2.Rep1: alpha=0.100, lambda=0.07781
## - Fold2.Rep1: alpha=0.100, lambda=0.07781
## + Fold2.Rep1: alpha=0.325, lambda=0.07781
## - Fold2.Rep1: alpha=0.325, lambda=0.07781
## + Fold2.Rep1: alpha=0.550, lambda=0.07781
## - Fold2.Rep1: alpha=0.550, lambda=0.07781
## + Fold2.Rep1: alpha=0.775, lambda=0.07781
## - Fold2.Rep1: alpha=0.775, lambda=0.07781
## + Fold2.Rep1: alpha=1.000, lambda=0.07781
## - Fold2.Rep1: alpha=1.000, lambda=0.07781
## + Fold3.Rep1: alpha=0.100, lambda=0.07781
## - Fold3.Rep1: alpha=0.100, lambda=0.07781
## + Fold3.Rep1: alpha=0.325, lambda=0.07781
## - Fold3.Rep1: alpha=0.325, lambda=0.07781
## + Fold3.Rep1: alpha=0.550, lambda=0.07781
## - Fold3.Rep1: alpha=0.550, lambda=0.07781
## + Fold3.Rep1: alpha=0.775, lambda=0.07781
## - Fold3.Rep1: alpha=0.775, lambda=0.07781
## + Fold3.Rep1: alpha=1.000, lambda=0.07781
## - Fold3.Rep1: alpha=1.000, lambda=0.07781
## + Fold1.Rep2: alpha=0.100, lambda=0.07781
## - Fold1.Rep2: alpha=0.100, lambda=0.07781
## + Fold1.Rep2: alpha=0.325, lambda=0.07781
## - Fold1.Rep2: alpha=0.325, lambda=0.07781
## + Fold1.Rep2: alpha=0.550, lambda=0.07781
## - Fold1.Rep2: alpha=0.550, lambda=0.07781
## + Fold1.Rep2: alpha=0.775, lambda=0.07781
## - Fold1.Rep2: alpha=0.775, lambda=0.07781
## + Fold1.Rep2: alpha=1.000, lambda=0.07781
## - Fold1.Rep2: alpha=1.000, lambda=0.07781
## + Fold2.Rep2: alpha=0.100, lambda=0.07781
## - Fold2.Rep2: alpha=0.100, lambda=0.07781
## + Fold2.Rep2: alpha=0.325, lambda=0.07781
## - Fold2.Rep2: alpha=0.325, lambda=0.07781
## + Fold2.Rep2: alpha=0.550, lambda=0.07781
## - Fold2.Rep2: alpha=0.550, lambda=0.07781
## + Fold2.Rep2: alpha=0.775, lambda=0.07781
## - Fold2.Rep2: alpha=0.775, lambda=0.07781
## + Fold2.Rep2: alpha=1.000, lambda=0.07781
## - Fold2.Rep2: alpha=1.000, lambda=0.07781
## + Fold3.Rep2: alpha=0.100, lambda=0.07781
## - Fold3.Rep2: alpha=0.100, lambda=0.07781
## + Fold3.Rep2: alpha=0.325, lambda=0.07781
## - Fold3.Rep2: alpha=0.325, lambda=0.07781
## + Fold3.Rep2: alpha=0.550, lambda=0.07781
## - Fold3.Rep2: alpha=0.550, lambda=0.07781
## + Fold3.Rep2: alpha=0.775, lambda=0.07781
## - Fold3.Rep2: alpha=0.775, lambda=0.07781
## + Fold3.Rep2: alpha=1.000, lambda=0.07781
## - Fold3.Rep2: alpha=1.000, lambda=0.07781
## + Fold1.Rep3: alpha=0.100, lambda=0.07781
## - Fold1.Rep3: alpha=0.100, lambda=0.07781
## + Fold1.Rep3: alpha=0.325, lambda=0.07781
## - Fold1.Rep3: alpha=0.325, lambda=0.07781
## + Fold1.Rep3: alpha=0.550, lambda=0.07781
## - Fold1.Rep3: alpha=0.550, lambda=0.07781
## + Fold1.Rep3: alpha=0.775, lambda=0.07781
## - Fold1.Rep3: alpha=0.775, lambda=0.07781
## + Fold1.Rep3: alpha=1.000, lambda=0.07781
## - Fold1.Rep3: alpha=1.000, lambda=0.07781
## + Fold2.Rep3: alpha=0.100, lambda=0.07781
## - Fold2.Rep3: alpha=0.100, lambda=0.07781
## + Fold2.Rep3: alpha=0.325, lambda=0.07781
## - Fold2.Rep3: alpha=0.325, lambda=0.07781
## + Fold2.Rep3: alpha=0.550, lambda=0.07781
## - Fold2.Rep3: alpha=0.550, lambda=0.07781
## + Fold2.Rep3: alpha=0.775, lambda=0.07781
## - Fold2.Rep3: alpha=0.775, lambda=0.07781
## + Fold2.Rep3: alpha=1.000, lambda=0.07781
## - Fold2.Rep3: alpha=1.000, lambda=0.07781
## + Fold3.Rep3: alpha=0.100, lambda=0.07781
## - Fold3.Rep3: alpha=0.100, lambda=0.07781
## + Fold3.Rep3: alpha=0.325, lambda=0.07781
## - Fold3.Rep3: alpha=0.325, lambda=0.07781
## + Fold3.Rep3: alpha=0.550, lambda=0.07781
## - Fold3.Rep3: alpha=0.550, lambda=0.07781
## + Fold3.Rep3: alpha=0.775, lambda=0.07781
## - Fold3.Rep3: alpha=0.775, lambda=0.07781
## + Fold3.Rep3: alpha=1.000, lambda=0.07781
## - Fold3.Rep3: alpha=1.000, lambda=0.07781
## Aggregating results
## Selecting tuning parameters
## Fitting alpha = 0.325, lambda = 0.0168 on full training set
## Length Class Mode
## a0 93 -none- numeric
## beta 2232 dgCMatrix S4
## df 93 -none- numeric
## dim 2 -none- numeric
## lambda 93 -none- numeric
## dev.ratio 93 -none- numeric
## nulldev 1 -none- numeric
## npasses 1 -none- numeric
## jerr 1 -none- numeric
## offset 1 -none- logical
## classnames 2 -none- character
## call 5 -none- call
## nobs 1 -none- numeric
## lambdaOpt 1 -none- numeric
## xNames 24 -none- character
## problemType 1 -none- character
## tuneValue 2 data.frame list
## obsLevels 2 -none- character
## [1] "min lambda > lambdaOpt:"
## (Intercept)
## -5.01773303
## NDSSName.my.fctr#Multimedia#
## -0.31536157
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.51314492
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 2.67877297
## NDSSName.my.fctr#U.S.#Education
## -0.82284989
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.02723099
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.37946894
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.34578807
## NDSSName.my.fctrBusiness#Technology#
## 0.31843410
## NDSSName.my.fctrForeign#World#
## -0.27776127
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.81472898
## NDSSName.my.fctrOpEd#Opinion#
## 3.35163474
## NDSSName.my.fctrScience#Health#
## 2.48264645
## NDSSName.my.fctrStyles##Fashion
## -0.49241325
## NDSSName.my.fctrStyles#U.S.#
## 2.22613995
## NDSSName.my.fctrTStyle##
## -0.72232021
## NDSSName.my.fctrTravel#Travel#
## -0.07844427
## NDSSName.my.fctrmyOther
## -0.22253379
## WordCount.log1p
## 0.26612836
## WordCount.root2
## 0.04648425
## [1] "max lambda < lambdaOpt:"
## (Intercept)
## -5.09261870
## NDSSName.my.fctr#Multimedia#
## -0.37993993
## NDSSName.my.fctr#Opinion#RoomForDebate
## -1.62062746
## NDSSName.my.fctr#Opinion#ThePublicEditor
## 2.72614928
## NDSSName.my.fctr#U.S.#Education
## -0.89871970
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook
## -0.04679393
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness
## -0.43224773
## NDSSName.my.fctrBusiness#Crosswords/Games#
## 3.38410449
## NDSSName.my.fctrBusiness#Technology#
## 0.33911630
## NDSSName.my.fctrForeign#World#
## -0.34241871
## NDSSName.my.fctrForeign#World#AsiaPacific
## -0.88251852
## NDSSName.my.fctrOpEd#Opinion#
## 3.38386039
## NDSSName.my.fctrScience#Health#
## 2.50486732
## NDSSName.my.fctrStyles##Fashion
## -0.56144626
## NDSSName.my.fctrStyles#U.S.#
## 2.24594116
## NDSSName.my.fctrTStyle##
## -0.77459454
## NDSSName.my.fctrTravel#Travel#
## -0.13420086
## NDSSName.my.fctrmyOther
## -0.31775172
## WordCount.log1p
## 0.27449523
## WordCount.root2
## 0.04746532
## Popular.fctr Popular.fctr.predict.Final.glmnet.N
## 1 N 5141
## 2 Y 338
## Popular.fctr.predict.Final.glmnet.Y
## 1 298
## 2 755
## Prediction
## Reference N Y
## N 5141 298
## Y 338 755
## Accuracy Kappa AccuracyLower AccuracyUpper AccuracyNull
## 9.026332e-01 6.454065e-01 8.951865e-01 9.097181e-01 8.326699e-01
## AccuracyPValue McnemarPValue
## 3.401653e-59 1.219958e-01
## id
## 1 Final.glmnet
## feats
## 1 WordCount.root2,WordCount.log1p,NDSSName.my.fctr,.rnorm,WordCount.nexp
## max.nTuningRuns min.elapsedtime.everything min.elapsedtime.final
## 1 25 19.047 0.389
## max.AUCpROC.fit max.Sens.fit max.Spec.fit max.AUCROCR.fit
## 1 0.7945167 0.9641478 0.6248856 0.9310545
## opt.prob.threshold.fit max.f.score.fit max.Accuracy.fit
## 1 0.3 0.7036347 0.907124
## max.AccuracyLower.fit max.AccuracyUpper.fit max.Kappa.fit
## 1 0.8951865 0.9097181 0.6380888
## max.AccuracySD.fit max.KappaSD.fit
## 1 0.004657963 0.02195699rm(ret_lst)
glb_chunks_df <- myadd_chunk(glb_chunks_df, "fit.data.training", major.inc=FALSE)## label step_major step_minor label_minor bgn end
## 14 fit.data.training 7 0 0 181.344 222.039
## 15 fit.data.training 7 1 1 222.039 NA
## elapsed
## 14 40.695
## 15 NA#stop(here"); glb_to_sav()
if (glb_is_classification && glb_is_binomial)
prob_threshold <- glb_models_df[glb_models_df$id == glb_sel_mdl_id,
"opt.prob.threshold.OOB"] else
prob_threshold <- NULL
if (grepl("Ensemble", glb_fin_mdl_id)) {
# Get predictions for each model in ensemble; Outliers that have been moved to OOB might not have been predicted yet
mdlEnsembleComps <- unlist(str_split(subset(glb_models_df,
id == glb_fin_mdl_id)$feats, ","))
if (glb_is_classification && glb_is_binomial)
mdlEnsembleComps <- gsub("\\.prob$", "", mdlEnsembleComps)
mdlEnsembleComps <- gsub(paste0("^",
gsub(".", "\\.", glb_rsp_var_out, fixed = TRUE)),
"", mdlEnsembleComps)
for (mdl_id in mdlEnsembleComps) {
glbObsTrn <- glb_get_predictions(df=glbObsTrn, mdl_id=mdl_id,
rsp_var_out=glb_rsp_var_out,
prob_threshold_def=prob_threshold)
glbObsNew <- glb_get_predictions(df=glbObsNew, mdl_id=mdl_id,
rsp_var_out=glb_rsp_var_out,
prob_threshold_def=prob_threshold)
}
}
glbObsTrn <- glb_get_predictions(df=glbObsTrn, mdl_id=glb_fin_mdl_id,
rsp_var_out=glb_rsp_var_out,
prob_threshold_def=prob_threshold)## Warning in glb_get_predictions(df = glbObsTrn, mdl_id = glb_fin_mdl_id, :
## Using default probability threshold: 0.1glb_featsimp_df <- myget_feats_importance(mdl=glb_fin_mdl,
featsimp_df=glb_featsimp_df)
glb_featsimp_df[, paste0(glb_fin_mdl_id, ".importance")] <- glb_featsimp_df$importance
print(glb_featsimp_df)## All.X.glmnet.importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 98.38173
## NDSSName.my.fctr#Opinion#ThePublicEditor 87.34197
## NDSSName.my.fctrScience#Health# 84.05064
## NDSSName.my.fctrStyles#U.S.# 80.66804
## NDSSName.my.fctrBusiness#Technology# 43.29623
## WordCount.log1p 34.01597
## WordCount.root2 32.29795
## .rnorm 31.33944
## NDSSName.my.fctrCulture## 31.33944
## NDSSName.my.fctrCulture#Arts# 31.33944
## NDSSName.my.fctrMetro#N.Y./Region# 31.33944
## WordCount.nexp 31.33944
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.33944
## NDSSName.my.fctrTravel#Travel# 31.31570
## NDSSName.my.fctrmyOther 27.85141
## NDSSName.my.fctrForeign#World# 30.00852
## NDSSName.my.fctr#Multimedia# 28.91940
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 27.78546
## NDSSName.my.fctrStyles##Fashion 22.02991
## NDSSName.my.fctrTStyle## 17.36032
## NDSSName.my.fctrForeign#World#AsiaPacific 19.29994
## NDSSName.my.fctr#U.S.#Education 18.25900
## NDSSName.my.fctr#Opinion#RoomForDebate 0.00000
## importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 99.96082
## NDSSName.my.fctr#Opinion#ThePublicEditor 86.61487
## NDSSName.my.fctrScience#Health# 82.33065
## NDSSName.my.fctrStyles#U.S.# 77.12194
## NDSSName.my.fctrBusiness#Technology# 38.62774
## WordCount.log1p 37.41260
## WordCount.root2 32.88366
## .rnorm 31.93272
## NDSSName.my.fctrCulture## 31.93272
## NDSSName.my.fctrCulture#Arts# 31.93272
## NDSSName.my.fctrMetro#N.Y./Region# 31.93272
## WordCount.nexp 31.93272
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.12990
## NDSSName.my.fctrTravel#Travel# 29.62778
## NDSSName.my.fctrmyOther 26.20874
## NDSSName.my.fctrForeign#World# 25.48953
## NDSSName.my.fctr#Multimedia# 24.73162
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 23.59037
## NDSSName.my.fctrStyles##Fashion 21.10031
## NDSSName.my.fctrTStyle## 16.67670
## NDSSName.my.fctrForeign#World#AsiaPacific 14.61070
## NDSSName.my.fctr#U.S.#Education 14.34221
## NDSSName.my.fctr#Opinion#RoomForDebate 0.00000
## Final.glmnet.importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 99.96082
## NDSSName.my.fctr#Opinion#ThePublicEditor 86.61487
## NDSSName.my.fctrScience#Health# 82.33065
## NDSSName.my.fctrStyles#U.S.# 77.12194
## NDSSName.my.fctrBusiness#Technology# 38.62774
## WordCount.log1p 37.41260
## WordCount.root2 32.88366
## .rnorm 31.93272
## NDSSName.my.fctrCulture## 31.93272
## NDSSName.my.fctrCulture#Arts# 31.93272
## NDSSName.my.fctrMetro#N.Y./Region# 31.93272
## WordCount.nexp 31.93272
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.12990
## NDSSName.my.fctrTravel#Travel# 29.62778
## NDSSName.my.fctrmyOther 26.20874
## NDSSName.my.fctrForeign#World# 25.48953
## NDSSName.my.fctr#Multimedia# 24.73162
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 23.59037
## NDSSName.my.fctrStyles##Fashion 21.10031
## NDSSName.my.fctrTStyle## 16.67670
## NDSSName.my.fctrForeign#World#AsiaPacific 14.61070
## NDSSName.my.fctr#U.S.#Education 14.34221
## NDSSName.my.fctr#Opinion#RoomForDebate 0.00000if (glb_is_classification && glb_is_binomial)
glb_analytics_diag_plots(obs_df=glbObsTrn, mdl_id=glb_fin_mdl_id,
prob_threshold=glb_models_df[glb_models_df$id == glb_sel_mdl_id,
"opt.prob.threshold.OOB"]) else
glb_analytics_diag_plots(obs_df=glbObsTrn, mdl_id=glb_fin_mdl_id) 
## [1] "Min/Max Boundaries: "
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 96 96 N NA
## 6370 6370 Y 0.9608409
## Popular.fctr.predict.All.X.glmnet
## 96 <NA>
## 6370 Y
## Popular.fctr.predict.All.X.glmnet.err
## 96 NA
## 6370 FALSE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 96 NA
## 6370 0.03915913
## Popular.fctr.predict.All.X.glmnet.accurate
## 96 NA
## 6370 TRUE
## Popular.fctr.predict.Final.glmnet.prob
## 96 0.006026625
## 6370 0.915289043
## Popular.fctr.predict.Final.glmnet
## 96 N
## 6370 Y
## Popular.fctr.predict.Final.glmnet.err
## 96 FALSE
## 6370 FALSE
## Popular.fctr.predict.Final.glmnet.err.abs
## 96 0.006026625
## 6370 0.084710957
## Popular.fctr.predict.Final.glmnet.accurate
## 96 TRUE
## 6370 TRUE
## Popular.fctr.predict.Final.glmnet.error .label
## 96 0 96
## 6370 0 6370
## [1] "Inaccurate: "
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 2182 2182 Y 0.006633964
## 4020 4020 Y NA
## 1696 1696 Y 0.012803835
## 4775 4775 Y NA
## 4721 4721 Y 0.020278808
## 364 364 Y 0.019247847
## Popular.fctr.predict.All.X.glmnet
## 2182 N
## 4020 <NA>
## 1696 N
## 4775 <NA>
## 4721 N
## 364 N
## Popular.fctr.predict.All.X.glmnet.err
## 2182 TRUE
## 4020 NA
## 1696 TRUE
## 4775 NA
## 4721 TRUE
## 364 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 2182 0.9933660
## 4020 NA
## 1696 0.9871962
## 4775 NA
## 4721 0.9797212
## 364 0.9807522
## Popular.fctr.predict.All.X.glmnet.accurate
## 2182 FALSE
## 4020 NA
## 1696 FALSE
## 4775 NA
## 4721 FALSE
## 364 FALSE
## Popular.fctr.predict.Final.glmnet.prob
## 2182 0.01576970
## 4020 0.02182292
## 1696 0.02432199
## 4775 0.03213481
## 4721 0.03466036
## 364 0.03629580
## Popular.fctr.predict.Final.glmnet
## 2182 N
## 4020 N
## 1696 N
## 4775 N
## 4721 N
## 364 N
## Popular.fctr.predict.Final.glmnet.err
## 2182 TRUE
## 4020 TRUE
## 1696 TRUE
## 4775 TRUE
## 4721 TRUE
## 364 TRUE
## Popular.fctr.predict.Final.glmnet.err.abs
## 2182 0.9842303
## 4020 0.9781771
## 1696 0.9756780
## 4775 0.9678652
## 4721 0.9653396
## 364 0.9637042
## Popular.fctr.predict.Final.glmnet.accurate
## 2182 FALSE
## 4020 FALSE
## 1696 FALSE
## 4775 FALSE
## 4721 FALSE
## 364 FALSE
## Popular.fctr.predict.Final.glmnet.error
## 2182 -0.08423030
## 4020 -0.07817708
## 1696 -0.07567801
## 4775 -0.06786519
## 4721 -0.06533964
## 364 -0.06370420
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 3029 3029 N NA
## 323 323 N 0.07847085
## 1181 1181 N NA
## 2064 2064 N 0.26251870
## 5287 5287 N 0.60379585
## 4786 4786 N 0.91965114
## Popular.fctr.predict.All.X.glmnet
## 3029 <NA>
## 323 N
## 1181 <NA>
## 2064 Y
## 5287 Y
## 4786 Y
## Popular.fctr.predict.All.X.glmnet.err
## 3029 NA
## 323 FALSE
## 1181 NA
## 2064 TRUE
## 5287 TRUE
## 4786 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 3029 NA
## 323 0.07847085
## 1181 NA
## 2064 0.26251870
## 5287 0.60379585
## 4786 0.91965114
## Popular.fctr.predict.All.X.glmnet.accurate
## 3029 NA
## 323 TRUE
## 1181 NA
## 2064 FALSE
## 5287 FALSE
## 4786 FALSE
## Popular.fctr.predict.Final.glmnet.prob
## 3029 0.1051303
## 323 0.1168606
## 1181 0.1702228
## 2064 0.2950796
## 5287 0.4139998
## 4786 0.8143734
## Popular.fctr.predict.Final.glmnet
## 3029 Y
## 323 Y
## 1181 Y
## 2064 Y
## 5287 Y
## 4786 Y
## Popular.fctr.predict.Final.glmnet.err
## 3029 TRUE
## 323 TRUE
## 1181 TRUE
## 2064 TRUE
## 5287 TRUE
## 4786 TRUE
## Popular.fctr.predict.Final.glmnet.err.abs
## 3029 0.1051303
## 323 0.1168606
## 1181 0.1702228
## 2064 0.2950796
## 5287 0.4139998
## 4786 0.8143734
## Popular.fctr.predict.Final.glmnet.accurate
## 3029 FALSE
## 323 FALSE
## 1181 FALSE
## 2064 FALSE
## 5287 FALSE
## 4786 FALSE
## Popular.fctr.predict.Final.glmnet.error
## 3029 0.005130305
## 323 0.016860557
## 1181 0.070222845
## 2064 0.195079625
## 5287 0.313999830
## 4786 0.714373384
## UniqueID Popular.fctr Popular.fctr.predict.All.X.glmnet.prob
## 221 221 N NA
## 3590 3590 N NA
## 472 472 N NA
## 2995 2995 N NA
## 6276 6276 N NA
## 3258 3258 N 0.9735424
## Popular.fctr.predict.All.X.glmnet
## 221 <NA>
## 3590 <NA>
## 472 <NA>
## 2995 <NA>
## 6276 <NA>
## 3258 Y
## Popular.fctr.predict.All.X.glmnet.err
## 221 NA
## 3590 NA
## 472 NA
## 2995 NA
## 6276 NA
## 3258 TRUE
## Popular.fctr.predict.All.X.glmnet.err.abs
## 221 NA
## 3590 NA
## 472 NA
## 2995 NA
## 6276 NA
## 3258 0.9735424
## Popular.fctr.predict.All.X.glmnet.accurate
## 221 NA
## 3590 NA
## 472 NA
## 2995 NA
## 6276 NA
## 3258 FALSE
## Popular.fctr.predict.Final.glmnet.prob
## 221 0.9169618
## 3590 0.9171703
## 472 0.9195159
## 2995 0.9203097
## 6276 0.9215804
## 3258 0.9245769
## Popular.fctr.predict.Final.glmnet
## 221 Y
## 3590 Y
## 472 Y
## 2995 Y
## 6276 Y
## 3258 Y
## Popular.fctr.predict.Final.glmnet.err
## 221 TRUE
## 3590 TRUE
## 472 TRUE
## 2995 TRUE
## 6276 TRUE
## 3258 TRUE
## Popular.fctr.predict.Final.glmnet.err.abs
## 221 0.9169618
## 3590 0.9171703
## 472 0.9195159
## 2995 0.9203097
## 6276 0.9215804
## 3258 0.9245769
## Popular.fctr.predict.Final.glmnet.accurate
## 221 FALSE
## 3590 FALSE
## 472 FALSE
## 2995 FALSE
## 6276 FALSE
## 3258 FALSE
## Popular.fctr.predict.Final.glmnet.error
## 221 0.8169618
## 3590 0.8171703
## 472 0.8195159
## 2995 0.8203097
## 6276 0.8215804
## 3258 0.8245769
dsp_feats_vctr <- c(NULL)
for(var in grep(".importance", names(glb_feats_df), fixed=TRUE, value=TRUE))
dsp_feats_vctr <- union(dsp_feats_vctr,
glb_feats_df[!is.na(glb_feats_df[, var]), "id"])
# print(glbObsTrn[glbObsTrn$UniqueID %in% FN_OOB_ids,
# grep(glb_rsp_var, names(glbObsTrn), value=TRUE)])
print(setdiff(names(glbObsTrn), names(glbObsAll)))## [1] "Popular.fctr.predict.Final.glmnet.prob"
## [2] "Popular.fctr.predict.Final.glmnet"
## [3] "Popular.fctr.predict.Final.glmnet.err"
## [4] "Popular.fctr.predict.Final.glmnet.err.abs"
## [5] "Popular.fctr.predict.Final.glmnet.accurate"for (col in setdiff(names(glbObsTrn), names(glbObsAll)))
# Merge or cbind ?
glbObsAll[glbObsAll$.src == "Train", col] <- glbObsTrn[, col]
print(setdiff(names(glbObsFit), names(glbObsAll)))## character(0)print(setdiff(names(glbObsOOB), names(glbObsAll)))## character(0)for (col in setdiff(names(glbObsOOB), names(glbObsAll)))
# Merge or cbind ?
glbObsAll[glbObsAll$.lcn == "OOB", col] <- glbObsOOB[, col]
print(setdiff(names(glbObsNew), names(glbObsAll)))## character(0)if (glb_save_envir)
save(glb_feats_df, glbObsAll,
#glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
glb_models_df, dsp_models_df, glb_models_lst, glb_model_type,
glb_sel_mdl, glb_sel_mdl_id,
glb_fin_mdl, glb_fin_mdl_id,
file=paste0(glb_out_pfx, "dsk.RData"))
replay.petrisim(pn=glb_analytics_pn,
replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
"data.training.all.prediction","model.final")), flip_coord=TRUE)## time trans "bgn " "fit.data.training.all " "predict.data.new " "end "
## 0.0000 multiple enabled transitions: data.training.all data.new model.selected firing: data.training.all
## 1.0000 1 2 1 0 0
## 1.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction firing: data.new
## 2.0000 2 1 1 1 0
## 2.0000 multiple enabled transitions: data.training.all data.new model.selected model.final data.training.all.prediction data.new.prediction firing: model.selected
## 3.0000 3 0 2 1 0
## 3.0000 multiple enabled transitions: model.final data.training.all.prediction data.new.prediction firing: data.training.all.prediction
## 4.0000 5 0 1 1 1
## 4.0000 multiple enabled transitions: model.final data.training.all.prediction data.new.prediction firing: model.final
## 5.0000 4 0 0 2 1
glb_chunks_df <- myadd_chunk(glb_chunks_df, "predict.data.new", major.inc=TRUE)## label step_major step_minor label_minor bgn end
## 15 fit.data.training 7 1 1 222.039 233.199
## 16 predict.data.new 8 0 0 233.199 NA
## elapsed
## 15 11.16
## 16 NAStep 8.0: predict data new
# Compute final model predictions
#glb_to_sav(); all.equal(sav_allobs_df, glbObsAll); all.equal(sav_trnobs_df, glbObsTrn); all.equal(sav_newobs_df, glbObsNew)
if (glb_is_classification && glb_is_binomial)
prob_threshold_def <-
glb_models_df[glb_models_df$id == glb_sel_mdl_id, "opt.prob.threshold.OOB"] else
prob_threshold_def <- NULL
for (obsSet in c("trn", "new")) {
obs_df <- switch(obsSet, all = glbObsAll, trn = glbObsTrn, new = glbObsNew)
obs_df <- glb_get_predictions(obs_df, mdl_id = glb_fin_mdl_id,
rsp_var_out = glb_rsp_var_out, prob_threshold_def = prob_threshold_def)
if (obsSet == "all") glbObsAll <- obs_df else
if (obsSet == "trn") glbObsTrn <- obs_df else
if (obsSet == "new") glbObsNew <- obs_df
}## Warning in glb_get_predictions(obs_df, mdl_id = glb_fin_mdl_id, rsp_var_out
## = glb_rsp_var_out, : Using default probability threshold: 0.1## Warning in glb_get_predictions(obs_df, mdl_id = glb_fin_mdl_id, rsp_var_out
## = glb_rsp_var_out, : Using default probability threshold: 0.1rm(obs_df)
glbObsAll <- orderBy(reformulate(glb_id_var), myrbind_df(glbObsTrn, glbObsNew))
glb_analytics_diag_plots(obs_df = glbObsNew, mdl_id = glb_fin_mdl_id,
prob_threshold = prob_threshold_def)## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## Warning: Removed 1870 rows containing missing values (geom_point).## Warning: Removed 1870 rows containing missing values (geom_point).
## NULLif (is.null(glb_out_obs)) obs_df <- glbObsNew else
obs_df <- switch(glb_out_obs,
all = glbObsAll, trn = glbObsTrn, new = glbObsNew)
require(stringr)## Loading required package: stringrif (glb_is_classification && glb_is_binomial) {
obsout_df <- obs_df[, glb_id_var, FALSE]
for (clmn in names(glb_out_vars_lst))
if (!grepl("^%<d-%", glb_out_vars_lst[[clmn]]))
obsout_df[, clmn] <- obs_df[, glb_out_vars_lst[[clmn]]] else {
feat <- str_trim(unlist(strsplit(glb_out_vars_lst[[clmn]], "%<d-%"))[2])
obsout_df[, clmn] <- obs_df[, eval(parse(text = feat))]
}
# glb_force_prediction_lst <- list()
# glb_force_prediction_lst[["0"]] <- c(11885, 11907, 11932, 11943,
# 12050, 12115, 12171,
# 12253, 12285, 12367, 12388, 12399,
# 12585)
# for (obs_id in glb_force_prediction_lst[["0"]]) {
# if (sum(glbObsAll[, glb_id_var] == obs_id) == 0)
# next
# if (is.na(glbObsAll[glbObsAll[, glb_id_var] == obs_id, ".grpid"]))
# stop(".grpid is NA")
# # submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] <-
# # max(0, submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] - 0.5)
# }
# glb_force_prediction_lst[["1"]] <- c(11871, 11875, 11886,
# 11913, 11931, 11937, 11967, 11982, 11990, 11991, 11994, 11999,
# 12000, 12002, 12004, 12018, 12021, 12065, 12072,
# 12111, 12114, 12126, 12134, 12152, 12172,
# 12213, 12214, 12233, 12265, 12278, 12299,
# 12446, 12491,
# 12505, 12576, 12608, 12630)
# for (obs_id in glb_force_prediction_lst[["1"]]) {
# if (sum(glbObsAll[, glb_id_var] == obs_id) == 0)
# next
# if (is.na(glbObsAll[glbObsAll[, glb_id_var] == obs_id, ".grpid"]))
# stop(".grpid is NA")
# # submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] <-
# # min(0.9999, submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] + 0.5)
# }
# rsp_var_out <- paste0(glb_rsp_var_out, glb_fin_mdl_id)
# for (obs_id in glbObsNew[!is.na(glbObsNew[, rsp_var_out]) &
# (glbObsNew[, rsp_var_out] == "Y") &
# (glbObsNew[ , "startprice"] > 675), "UniqueID"]) {
# # submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] <-
# # max(0, submit_df[submit_df[, glb_id_var] == obs_id, "Probability1"] - 0.5)
# }
} else {
# submit_df <- glbObsNew[, c(glb_id_var,
# paste0(glb_rsp_var_out, glb_fin_mdl_id))]
obsout_df <- obs_df[, glb_id_var, FALSE]
for (clmn in names(glb_out_vars_lst))
if (!grepl("^%<d-%", glb_out_vars_lst[[clmn]]))
obsout_df[, clmn] <- obs_df[, glb_out_vars_lst[[clmn]]] else {
feat <- str_trim(unlist(strsplit(glb_out_vars_lst[[clmn]], "%<d-%"))[2])
obsout_df[, clmn] <- obs_df[, eval(parse(text=feat))]
}
}
if (glb_is_classification) {
rsp_var_out <- paste0(glb_rsp_var_out, glb_fin_mdl_id)
if (".grpid" %in% names(glbObsNew)) {
# Dups were found in glbObsAll
tmp_newobs_df <- subset(glbObsNew[, c(glb_id_var, ".grpid", rsp_var_out)],
!is.na(.grpid))
tmp_newobs_df <-
merge(tmp_newobs_df, dupgrps_df, by = ".grpid", all.x = TRUE)
tmp_newobs_df <-
merge(tmp_newobs_df, obsout_df, by = glb_id_var, all.x = TRUE)
tmp_newobs_df$.err <-
((tmp_newobs_df$Probability1 > 0.5) & (tmp_newobs_df$sold.0 > 0) |
(tmp_newobs_df$Probability1 < 0.5) & (tmp_newobs_df$sold.1 > 0))
tmp_newobs_df <- orderBy(~UniqueID, subset(tmp_newobs_df, .err == TRUE))
print(sprintf("Prediction errors in duplicates: %d", nrow(tmp_newobs_df)))
print(tmp_newobs_df)
}
tmp_newobs_df <- cbind(glbObsNew, obsout_df[, "Probability1", FALSE])
# Check predictions that are outside of data ranges
#stop(here")
require(stringr)
tmp_feats_df <- subset(glb_feats_df,
!nzv &
(exclude.as.feat != 1) &
!grepl(".fctr", id, fixed=TRUE))[, "id", FALSE]
ranges_all_df <- glbObsAll[, tmp_feats_df$id] %>%
dplyr::summarise_each(funs(min(., na.rm=TRUE),
max(., na.rm=TRUE))) %>%
tidyr::gather() %>%
dplyr::mutate(id=str_sub(key, 1, -5),
stat=str_sub(key, -3)) %>%
dplyr::select(-key) %>%
tidyr::spread(stat, value)
# sav_ranges_trn_df <- ranges_trn_df; all.equal(sav_ranges_trn_df, ranges_trn_df)
# sav_ranges_new_df <- ranges_new_df; all.equal(sav_ranges_new_df, ranges_new_df)
get_ranges_df <- function(obs_df, feats, class_var) {
require(tidyr)
ranges_df <- obs_df[, c(class_var, feats)] %>%
dplyr::group_by_(class_var) %>%
dplyr::summarise_each(funs(min(., na.rm=TRUE),
max(., na.rm=TRUE))) %>%
tidyr::gather(key, value, -1) %>%
mutate(id=str_sub(key, 1, -5),
stat.vname=paste0(str_sub(key, -3), ".", class_var)) %>%
unite_("stat.class", c("stat.vname", class_var), sep=".") %>%
dplyr::select(-key) %>%
spread(stat.class, value)
return(ranges_df)
}
rsp_var_out_OOB <- paste0(glb_rsp_var_out, glb_sel_mdl_id)
rsp_var_out_new <- paste0(glb_rsp_var_out, glb_fin_mdl_id)
ranges_trn_df <- get_ranges_df(obs_df=glbObsTrn, feats=tmp_feats_df$id,
class_var=glb_rsp_var)
ranges_fit_df <- get_ranges_df(obs_df=glbObsFit, feats=tmp_feats_df$id,
class_var=glb_rsp_var)
ranges_OOB_df <- get_ranges_df(obs_df=glbObsOOB, feats=tmp_feats_df$id,
class_var=rsp_var_out_OOB)
ranges_new_df <- get_ranges_df(obs_df=glbObsNew, feats=tmp_feats_df$id,
class_var=rsp_var_out_new)
for (obsset in c("OOB", "new")) {
if (obsset == "OOB") {
ranges_ref_df <- ranges_fit_df; obs_df <- glbObsOOB;
rsp_var_out_obs <- rsp_var_out_OOB; sprintf_pfx <- "OOBobs";
} else {
ranges_ref_df <- ranges_trn_df; obs_df <- glbObsNew;
rsp_var_out_obs <- rsp_var_out_new; sprintf_pfx <- "newobs";
}
plt_feats_df <- glb_feats_df %>%
merge(ranges_all_df, all=TRUE) %>%
merge(ranges_ref_df, all=TRUE) %>%
merge(ranges_OOB_df, all=TRUE) %>%
merge(ranges_new_df, all=TRUE) %>%
subset(!is.na(min) & (id != ".rnorm"))
row.names(plt_feats_df) <- plt_feats_df$id
range_outlier_ids <- c(NULL)
for (clss in unique(obs_df[, rsp_var_out_obs])) {
for (stat in c("min", "max")) {
if (stat == "min") {
dsp_feats <- plt_feats_df[
which(plt_feats_df[, paste("min", rsp_var_out_obs, clss, sep=".")] <
plt_feats_df[, paste("min", glb_rsp_var, clss, sep=".")]), "id"]
} else {
dsp_feats <- plt_feats_df[
which(plt_feats_df[, paste("max", rsp_var_out_obs, clss, sep=".")] >
plt_feats_df[, paste("max", glb_rsp_var, clss, sep=".")]), "id"]
}
if (length(dsp_feats) > 0) {
ths_ids <- c(NULL)
for (feat in dsp_feats) {
if (stat == "min") {
ths_ids <- union(ths_ids,
obs_df[(obs_df[, rsp_var_out_obs] == clss) &
(obs_df[, feat] <
plt_feats_df[plt_feats_df$id == feat, paste("min", glb_rsp_var, clss, sep=".")]),
glb_id_var])
} else {
ths_ids <- union(ths_ids,
obs_df[(obs_df[, rsp_var_out_obs] == clss) &
(obs_df[, feat] >
plt_feats_df[plt_feats_df$id == feat, paste("max", glb_rsp_var, clss, sep=".")]),
glb_id_var])
}
}
tmp_obs_df <- obs_df[obs_df[, glb_id_var] %in% ths_ids,
c(glb_id_var, rsp_var_out_obs, dsp_feats)]
if (stat == "min") {
print(sprintf("%s %s %s: min < min of Train range: %d",
sprintf_pfx, rsp_var_out_obs, clss, nrow(tmp_obs_df)))
} else {
print(sprintf("%s %s %s: max > max of Train range: %d",
sprintf_pfx, rsp_var_out_obs, clss, nrow(tmp_obs_df)))
}
myprint_df(tmp_obs_df)
print(subset(plt_feats_df, id %in% dsp_feats))
range_outlier_ids <- union(range_outlier_ids, ths_ids)
}
}
}
print(sprintf("%s total range outliers: %d", sprintf_pfx,
length(range_outlier_ids)))
}
}## Loading required package: tidyr
##
## Attaching package: 'tidyr'
##
## The following object is masked from 'package:Matrix':
##
## expand## [1] "OOBobs Popular.fctr.predict.All.X.glmnet Y: min < min of Train range: 1"
## UniqueID Popular.fctr.predict.All.X.glmnet WordCount.log1p
## 6435 6435 Y 0
## WordCount.root2
## 6435 0
## id cor.y exclude.as.feat cor.y.abs
## WordCount.log1p WordCount.log1p 0.2543196 FALSE 0.2543196
## WordCount.root2 WordCount.root2 0.2921207 FALSE 0.2921207
## cor.high.X freqRatio percentUnique zeroVar nzv
## WordCount.log1p WordCount.root2 2.315789 24.15799 FALSE FALSE
## WordCount.root2 <NA> 2.315789 24.15799 FALSE FALSE
## is.cor.y.abs.low interaction.feat shapiro.test.p.value
## WordCount.log1p FALSE NA 1.576866e-49
## WordCount.root2 FALSE NA 4.556481e-30
## rsp_var_raw id_var rsp_var max min
## WordCount.log1p FALSE NA NA 9.29771 0
## WordCount.root2 FALSE NA NA 104.46052 0
## max.Popular.fctr.N max.Popular.fctr.Y min.Popular.fctr.N
## WordCount.log1p 8.819665 9.29771 0
## WordCount.root2 82.249620 104.46052 0
## min.Popular.fctr.Y max.Popular.fctr.predict.All.X.glmnet.N
## WordCount.log1p 1.94591 8.117014
## WordCount.root2 2.44949 57.879185
## max.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.log1p 9.140883
## WordCount.root2 96.581572
## min.Popular.fctr.predict.All.X.glmnet.N
## WordCount.log1p 0
## WordCount.root2 0
## min.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.log1p 0
## WordCount.root2 0
## max.Popular.fctr.predict.Final.glmnet.N
## WordCount.log1p 7.797702
## WordCount.root2 49.335586
## max.Popular.fctr.predict.Final.glmnet.Y
## WordCount.log1p 8.692322
## WordCount.root2 77.175126
## min.Popular.fctr.predict.Final.glmnet.N
## WordCount.log1p 0
## WordCount.root2 0
## min.Popular.fctr.predict.Final.glmnet.Y
## WordCount.log1p 1.609438
## WordCount.root2 2.000000
## [1] "OOBobs Popular.fctr.predict.All.X.glmnet Y: max > max of Train range: 1"
## UniqueID Popular.fctr.predict.All.X.glmnet WordCount.nexp
## 6435 6435 Y 1
## id cor.y exclude.as.feat cor.y.abs
## WordCount.nexp WordCount.nexp -0.0532084 FALSE 0.0532084
## cor.high.X freqRatio percentUnique zeroVar nzv
## WordCount.nexp <NA> 17.76136 11.32884 FALSE FALSE
## is.cor.y.abs.low interaction.feat shapiro.test.p.value
## WordCount.nexp FALSE NA 9.108805e-94
## rsp_var_raw id_var rsp_var max min max.Popular.fctr.N
## WordCount.nexp FALSE NA NA 1 0 1
## max.Popular.fctr.Y min.Popular.fctr.N min.Popular.fctr.Y
## WordCount.nexp 0.002478752 0 0
## max.Popular.fctr.predict.All.X.glmnet.N
## WordCount.nexp 1
## max.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.nexp 1
## min.Popular.fctr.predict.All.X.glmnet.N
## WordCount.nexp 0
## min.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.nexp 0
## max.Popular.fctr.predict.Final.glmnet.N
## WordCount.nexp 1
## max.Popular.fctr.predict.Final.glmnet.Y
## WordCount.nexp 0.01831564
## min.Popular.fctr.predict.Final.glmnet.N
## WordCount.nexp 0
## min.Popular.fctr.predict.Final.glmnet.Y
## WordCount.nexp 0
## [1] "OOBobs total range outliers: 1"
## [1] "newobs Popular.fctr.predict.Final.glmnet Y: min < min of Train range: 1"
## UniqueID Popular.fctr.predict.Final.glmnet WordCount.log1p
## 8217 8217 Y 1.609438
## WordCount.root2
## 8217 2
## id cor.y exclude.as.feat cor.y.abs
## WordCount.log1p WordCount.log1p 0.2543196 FALSE 0.2543196
## WordCount.root2 WordCount.root2 0.2921207 FALSE 0.2921207
## cor.high.X freqRatio percentUnique zeroVar nzv
## WordCount.log1p WordCount.root2 2.315789 24.15799 FALSE FALSE
## WordCount.root2 <NA> 2.315789 24.15799 FALSE FALSE
## is.cor.y.abs.low interaction.feat shapiro.test.p.value
## WordCount.log1p FALSE NA 1.576866e-49
## WordCount.root2 FALSE NA 4.556481e-30
## rsp_var_raw id_var rsp_var max min
## WordCount.log1p FALSE NA NA 9.29771 0
## WordCount.root2 FALSE NA NA 104.46052 0
## max.Popular.fctr.N max.Popular.fctr.Y min.Popular.fctr.N
## WordCount.log1p 8.819665 9.29771 0
## WordCount.root2 82.249620 104.46052 0
## min.Popular.fctr.Y max.Popular.fctr.predict.All.X.glmnet.N
## WordCount.log1p 1.94591 8.117014
## WordCount.root2 2.44949 57.879185
## max.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.log1p 9.140883
## WordCount.root2 96.581572
## min.Popular.fctr.predict.All.X.glmnet.N
## WordCount.log1p 0
## WordCount.root2 0
## min.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.log1p 0
## WordCount.root2 0
## max.Popular.fctr.predict.Final.glmnet.N
## WordCount.log1p 7.797702
## WordCount.root2 49.335586
## max.Popular.fctr.predict.Final.glmnet.Y
## WordCount.log1p 8.692322
## WordCount.root2 77.175126
## min.Popular.fctr.predict.Final.glmnet.N
## WordCount.log1p 0
## WordCount.root2 0
## min.Popular.fctr.predict.Final.glmnet.Y
## WordCount.log1p 1.609438
## WordCount.root2 2.000000
## [1] "newobs Popular.fctr.predict.Final.glmnet Y: max > max of Train range: 1"
## UniqueID Popular.fctr.predict.Final.glmnet WordCount.nexp
## 8217 8217 Y 0.01831564
## id cor.y exclude.as.feat cor.y.abs
## WordCount.nexp WordCount.nexp -0.0532084 FALSE 0.0532084
## cor.high.X freqRatio percentUnique zeroVar nzv
## WordCount.nexp <NA> 17.76136 11.32884 FALSE FALSE
## is.cor.y.abs.low interaction.feat shapiro.test.p.value
## WordCount.nexp FALSE NA 9.108805e-94
## rsp_var_raw id_var rsp_var max min max.Popular.fctr.N
## WordCount.nexp FALSE NA NA 1 0 1
## max.Popular.fctr.Y min.Popular.fctr.N min.Popular.fctr.Y
## WordCount.nexp 0.002478752 0 0
## max.Popular.fctr.predict.All.X.glmnet.N
## WordCount.nexp 1
## max.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.nexp 1
## min.Popular.fctr.predict.All.X.glmnet.N
## WordCount.nexp 0
## min.Popular.fctr.predict.All.X.glmnet.Y
## WordCount.nexp 0
## max.Popular.fctr.predict.Final.glmnet.N
## WordCount.nexp 1
## max.Popular.fctr.predict.Final.glmnet.Y
## WordCount.nexp 0.01831564
## min.Popular.fctr.predict.Final.glmnet.N
## WordCount.nexp 0
## min.Popular.fctr.predict.Final.glmnet.Y
## WordCount.nexp 0
## [1] "newobs total range outliers: 1"#stop(here"); glb_to_sav(); sav_obsout_df <- obsout_df; all.equal(sav_obsout_df, obsout_df); obsout_df <- sav_obsout_df
# This does not work for classification since AUC distribution might be different for different models
# -> Run glm on .prob from this glb_fin_mdl_id & .prob from stacked file & stack condition as a feature
if (!is.null(glbOutStackFnames)) {
for (fname in glbOutStackFnames) {
print(sprintf("Stacking file %s to prediction output...", fname))
#obsout_df <- dplyr::arrange_(rbind(obsout_df, read.csv(fname)), "UniqueID")
obsout_df <- dplyr::arrange_(rbind(obsout_df,
#read.csv(fname) %>% filter(!(UniqueID %in% obsout_df$UniqueID))),
#read.csv(fname) %>% filter(!(UniqueID %in% obsout_df[, glb_id_var]))),
read.csv(fname) %>%
dplyr::filter_(interp(~!(var %in% obsout_df$var),
var = as.name(glb_id_var)))),
glb_id_var)
if (nrow(obsout_df) != length(unique(obsout_df[, glb_id_var])))
stop("Potential dups in stacked prediction output")
}
}
out_fname <- paste0(glb_out_pfx, "out.csv")
write.csv(obsout_df, out_fname, quote = FALSE, row.names = FALSE)
#cat(" ", "\n", file=submit_fn, append=TRUE)
# print(orderBy(~ -max.auc.OOB, glb_models_df[, c("model_id",
# "max.auc.OOB", "max.Accuracy.OOB")]))
for (txt_var in glbFeatsText) {
# Print post-stem-words but need post-stop-words for debugging ?
print(sprintf(" All post-stem-words TfIDf terms for %s:", txt_var))
myprint_df(glb_post_stem_words_terms_df_lst[[txt_var]])
TfIdf_mtrx <- glb_post_stem_words_TfIdf_mtrx_lst[[txt_var]]
print(glbObsAll[
which(TfIdf_mtrx[, tail(glb_post_stem_words_terms_df_lst[[txt_var]], 1)$pos] > 0),
c(glb_id_var, glbFeatsText)])
print(nrow(subset(glb_post_stem_words_terms_df_lst[[txt_var]], freq == 1)))
#print(glbObsAll[which(TfIdf_mtrx[, 207] > 0), c(glb_id_var, glbFeatsText)])
#unlist(strsplit(glbObsAll[2157, "description"], ""))
#glbObsAll[2442, c(glb_id_var, glbFeatsText)]
#TfIdf_mtrx[2442, TfIdf_mtrx[2442, ] > 0]
print(sprintf(" Top_n post_stem_words TfIDf terms for %s:", txt_var))
tmp_df <- glb_post_stem_words_terms_df_lst[[txt_var]]
top_n_vctr <- tmp_df$term[1:glb_txt_top_n[[txt_var]]]
tmp_freq1_df <- subset(tmp_df, freq == 1)
tmp_freq1_df$top_n <- grepl(paste0(top_n_vctr, collapse="|"), tmp_freq1_df$term)
print(subset(tmp_freq1_df, top_n == TRUE))
}
if (glb_is_classification && glb_is_binomial)
print(glb_models_df[glb_models_df$model_id == glb_sel_mdl_id,
"opt.prob.threshold.OOB"])## numeric(0)print(sprintf("glb_sel_mdl_id: %s", glb_sel_mdl_id))## [1] "glb_sel_mdl_id: All.X.glmnet"print(sprintf("glb_fin_mdl_id: %s", glb_fin_mdl_id))## [1] "glb_fin_mdl_id: Final.glmnet"print(dsp_models_df)## id max.Accuracy.OOB
## Max.cor.Y.rpart Max.cor.Y.rpart 0.8200231
## All.X.glmnet All.X.glmnet 0.7690972
## Max.cor.Y.rcv.1X1.cp.0.rpart Max.cor.Y.rcv.1X1.cp.0.rpart 0.7673611
## Max.cor.Y.rcv.1X1.glmnet Max.cor.Y.rcv.1X1.glmnet 0.7604167
## Max.cor.Y.rcv.5X3.glmnet Max.cor.Y.rcv.5X3.glmnet 0.7604167
## Max.cor.Y.rcv.5X1.glmnet Max.cor.Y.rcv.5X1.glmnet 0.7604167
## Max.cor.Y.rcv.5X5.glmnet Max.cor.Y.rcv.5X5.glmnet 0.7604167
## Max.cor.Y.rcv.3X1.glmnet Max.cor.Y.rcv.3X1.glmnet 0.7575231
## Max.cor.Y.rcv.3X3.glmnet Max.cor.Y.rcv.3X3.glmnet 0.7575231
## Interact.High.cor.Y.glmnet Interact.High.cor.Y.glmnet 0.7575231
## Low.cor.X.glmnet Low.cor.X.glmnet 0.7575231
## Max.cor.Y.rcv.3X5.glmnet Max.cor.Y.rcv.3X5.glmnet 0.7575231
## MFO.myMFO_classfr MFO.myMFO_classfr 0.1331019
## Random.myrandom_classfr Random.myrandom_classfr 0.1331019
## max.AUCROCR.OOB max.AUCpROC.OOB
## Max.cor.Y.rpart 0.5892132 0.5870523
## All.X.glmnet 0.8061009 0.5965780
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.7773858 0.6174697
## Max.cor.Y.rcv.1X1.glmnet 0.8116126 0.5962443
## Max.cor.Y.rcv.5X3.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.5X1.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.5X5.glmnet 0.8114863 0.5962443
## Max.cor.Y.rcv.3X1.glmnet 0.8067975 0.5962443
## Max.cor.Y.rcv.3X3.glmnet 0.8067975 0.5962443
## Interact.High.cor.Y.glmnet 0.8067975 0.5962443
## Low.cor.X.glmnet 0.8067975 0.5962443
## Max.cor.Y.rcv.3X5.glmnet 0.8067975 0.5962443
## MFO.myMFO_classfr 0.5000000 0.5000000
## Random.myrandom_classfr 0.4857956 0.5125675
## max.Accuracy.fit opt.prob.threshold.fit
## Max.cor.Y.rpart 0.9296422 0.6
## All.X.glmnet 0.9326952 0.4
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.9381765 0.4
## Max.cor.Y.rcv.1X1.glmnet 0.9329725 0.5
## Max.cor.Y.rcv.5X3.glmnet 0.9333905 0.5
## Max.cor.Y.rcv.5X1.glmnet 0.9331818 0.5
## Max.cor.Y.rcv.5X5.glmnet 0.9331816 0.5
## Max.cor.Y.rcv.3X1.glmnet 0.9335973 0.4
## Max.cor.Y.rcv.3X3.glmnet 0.9333193 0.4
## Interact.High.cor.Y.glmnet 0.9333193 0.4
## Low.cor.X.glmnet 0.9333193 0.4
## Max.cor.Y.rcv.3X5.glmnet 0.9332218 0.4
## MFO.myMFO_classfr 0.1796420 0.1
## Random.myrandom_classfr 0.1796420 0.1
## opt.prob.threshold.OOB
## Max.cor.Y.rpart 0.6
## All.X.glmnet 0.1
## Max.cor.Y.rcv.1X1.cp.0.rpart 0.1
## Max.cor.Y.rcv.1X1.glmnet 0.1
## Max.cor.Y.rcv.5X3.glmnet 0.1
## Max.cor.Y.rcv.5X1.glmnet 0.1
## Max.cor.Y.rcv.5X5.glmnet 0.1
## Max.cor.Y.rcv.3X1.glmnet 0.1
## Max.cor.Y.rcv.3X3.glmnet 0.1
## Interact.High.cor.Y.glmnet 0.1
## Low.cor.X.glmnet 0.1
## Max.cor.Y.rcv.3X5.glmnet 0.1
## MFO.myMFO_classfr 0.1
## Random.myrandom_classfr 0.1if (glb_is_regression) {
print(sprintf("%s OOB RMSE: %0.4f", glb_sel_mdl_id,
glb_models_df[glb_models_df$model_id == glb_sel_mdl_id, "min.RMSE.OOB"]))
if (!is.null(glb_category_var)) {
tmp_OOBobs_df <- glbObsOOB[, c(glb_category_var, glb_rsp_var,
predct_error_var_name)]
names(tmp_OOBobs_df)[length(names(tmp_OOBobs_df))] <- "error.abs.OOB"
sOOB_ctgry_df <- dplyr::group_by_(tmp_OOBobs_df, glb_category_var)
sOOB_ctgry_df <- dplyr::count(sOOB_ctgry_df,
startprice.log10.abs.OOB.sum = sum(abs(startprice.log10)),
err.abs.OOB.sum = sum(error.abs.OOB),
err.abs.OOB.mean = mean(error.abs.OOB))
names(sOOB_ctgry_df)[4] <- ".n.OOB"
sOOB_ctgry_df <- dplyr::ungroup(sOOB_ctgry_df)
#intersect(names(glb_ctgry_df), names(sOOB_ctgry_df))
glb_ctgry_df <- merge(glb_ctgry_df, sOOB_ctgry_df, all=TRUE)
print(orderBy(~-err.abs.OOB.mean, glb_ctgry_df))
}
if ((glb_rsp_var %in% names(glbObsNew)) &&
!(any(is.na(glbObsNew[, glb_rsp_var])))) {
pred_stats_df <-
mypredict_mdl(mdl=glb_models_lst[[glb_fin_mdl_id]],
df=glbObsNew,
rsp_var=glb_rsp_var,
rsp_var_out=glb_rsp_var_out,
model_id_method=glb_fin_mdl_id,
label="new",
model_summaryFunction=glb_sel_mdl$control$summaryFunction,
model_metric=glb_sel_mdl$metric,
model_metric_maximize=glb_sel_mdl$maximize,
ret_type="stats")
print(sprintf("%s prediction stats for glbObsNew:", glb_fin_mdl_id))
print(pred_stats_df)
}
}
if (glb_is_classification) {
print(sprintf("%s OOB confusion matrix & accuracy: ", glb_sel_mdl_id))
print(t(confusionMatrix(glbObsOOB[, paste0(glb_rsp_var_out, glb_sel_mdl_id)],
glbObsOOB[, glb_rsp_var])$table))
if (!is.null(glb_category_var)) {
glb_ctgry_df <- merge(glb_ctgry_df,
myget_category_stats(obs_df = glbObsTrn, mdl_id = glb_fin_mdl_id,
label = "trn"),
by = glb_category_var, all = TRUE)
row.names(glb_ctgry_df) <- glb_ctgry_df[, glb_category_var]
glb_ctgry_df <- merge(glb_ctgry_df,
myget_category_stats(obs_df = glbObsNew, mdl_id = glb_fin_mdl_id,
label="new"),
by = glb_category_var, all = TRUE)
row.names(glb_ctgry_df) <- glb_ctgry_df[, glb_category_var]
if (any(grepl("OOB", glbMdlMetricsEval)))
print(orderBy(~-err.abs.OOB.mean, glb_ctgry_df[, -1])) else
print(orderBy(~-err.abs.fit.mean, glb_ctgry_df[, -1]))
print(colSums(glb_ctgry_df[, -grep(glb_category_var,
names(glb_ctgry_df))]))
}
if ((glb_rsp_var %in% names(glbObsNew)) &&
!(any(is.na(glbObsNew[, glb_rsp_var])))) {
print(sprintf("%s new confusion matrix & accuracy: ", glb_fin_mdl_id))
print(t(confusionMatrix(
glbObsNew[, paste0(glb_rsp_var_out, glb_fin_mdl_id)],
glbObsNew[, glb_rsp_var])$table))
}
} ## [1] "All.X.glmnet OOB confusion matrix & accuracy: "
## Prediction
## Reference N Y
## N 1176 322
## Y 77 153
## .n.OOB .n.Fit .n.Tst .freqRatio.Fit
## OpEd#Opinion# 89 437 164 0.090965862
## Styles#U.S.# 50 127 61 0.026436303
## Science#Health# 48 148 57 0.030807660
## Business#Crosswords/Games# 18 105 42 0.021856786
## #Opinion#ThePublicEditor 4 16 10 0.003330558
## Business#Technology# 126 213 114 0.044338052
## Business#BusinessDay#Dealbook 323 629 304 0.130932556
## ## 371 913 342 0.190049958
## Metro#N.Y./Region# 70 128 67 0.026644463
## Culture#Arts# 185 490 174 0.101998335
## Business#BusinessDay#SmallBusiness 40 100 41 0.020815987
## Styles##Fashion 15 104 15 0.021648626
## #Opinion#RoomForDebate 20 42 20 0.008742714
## #Multimedia# 49 92 52 0.019150708
## Travel#Travel# 34 83 35 0.017277269
## TStyle## 101 623 105 0.129683597
## Foreign#World#AsiaPacific 53 150 56 0.031223980
## myOther 5 33 5 0.006869276
## Culture## 1 NA 70 NA
## #U.S.#Education 82 243 89 0.050582848
## Foreign#World# 44 128 47 0.026644463
## .freqRatio.OOB .freqRatio.Tst
## OpEd#Opinion# 0.0515046296 0.087700535
## Styles#U.S.# 0.0289351852 0.032620321
## Science#Health# 0.0277777778 0.030481283
## Business#Crosswords/Games# 0.0104166667 0.022459893
## #Opinion#ThePublicEditor 0.0023148148 0.005347594
## Business#Technology# 0.0729166667 0.060962567
## Business#BusinessDay#Dealbook 0.1869212963 0.162566845
## ## 0.2146990741 0.182887701
## Metro#N.Y./Region# 0.0405092593 0.035828877
## Culture#Arts# 0.1070601852 0.093048128
## Business#BusinessDay#SmallBusiness 0.0231481481 0.021925134
## Styles##Fashion 0.0086805556 0.008021390
## #Opinion#RoomForDebate 0.0115740741 0.010695187
## #Multimedia# 0.0283564815 0.027807487
## Travel#Travel# 0.0196759259 0.018716578
## TStyle## 0.0584490741 0.056149733
## Foreign#World#AsiaPacific 0.0306712963 0.029946524
## myOther 0.0028935185 0.002673797
## Culture## 0.0005787037 0.037433155
## #U.S.#Education 0.0474537037 0.047593583
## Foreign#World# 0.0254629630 0.025133690
## err.abs.fit.sum err.abs.fit.mean .n.fit
## OpEd#Opinion# 101.962286 0.23332331 437
## Styles#U.S.# 51.707342 0.40714443 127
## Science#Health# 48.666469 0.32882749 148
## Business#Crosswords/Games# 19.346059 0.18424818 105
## #Opinion#ThePublicEditor 4.044658 0.25279116 16
## Business#Technology# 46.446343 0.21805795 213
## Business#BusinessDay#Dealbook 76.983498 0.12239030 629
## ## 91.828632 0.10057901 913
## Metro#N.Y./Region# 14.444993 0.11285150 128
## Culture#Arts# 41.599488 0.08489691 490
## Business#BusinessDay#SmallBusiness 8.607239 0.08607239 100
## Styles##Fashion 3.033172 0.02916512 104
## #Opinion#RoomForDebate 1.938007 0.04614303 42
## #Multimedia# 4.739265 0.05151375 92
## Travel#Travel# 2.569132 0.03095340 83
## TStyle## 14.962143 0.02401628 623
## Foreign#World#AsiaPacific 7.489991 0.04993328 150
## myOther 1.720179 0.05212665 33
## Culture## NA NA NA
## #U.S.#Education 4.569704 0.01880537 243
## Foreign#World# 3.357443 0.02623002 128
## err.abs.OOB.sum err.abs.OOB.mean
## OpEd#Opinion# 52.11236621 0.58553220
## Styles#U.S.# 27.39861838 0.54797237
## Science#Health# 25.12604617 0.52345930
## Business#Crosswords/Games# 9.01410297 0.50078350
## #Opinion#ThePublicEditor 1.97530353 0.49382588
## Business#Technology# 28.55232120 0.22660572
## Business#BusinessDay#Dealbook 57.75141149 0.17879694
## ## 63.39672286 0.17088065
## Metro#N.Y./Region# 11.12676976 0.15895385
## Culture#Arts# 28.98681899 0.15668551
## Business#BusinessDay#SmallBusiness 3.84496527 0.09612413
## Styles##Fashion 1.42711318 0.09514088
## #Opinion#RoomForDebate 1.78910542 0.08945527
## #Multimedia# 2.85858104 0.05833839
## Travel#Travel# 1.97769164 0.05816740
## TStyle## 5.15834308 0.05107270
## Foreign#World#AsiaPacific 2.69695832 0.05088601
## myOther 0.23180569 0.04636114
## Culture## 0.02849447 0.02849447
## #U.S.#Education 2.22191005 0.02709646
## Foreign#World# 1.13916144 0.02589003
## err.abs.trn.sum err.abs.trn.mean .n.trn
## OpEd#Opinion# 189.20017098 0.35969614 526
## Styles#U.S.# 83.39117680 0.47113659 177
## Science#Health# 80.38081529 0.41010620 196
## Business#Crosswords/Games# 36.88161340 0.29985052 123
## #Opinion#ThePublicEditor 7.55559605 0.37777980 20
## Business#Technology# 76.10301242 0.22449266 339
## Business#BusinessDay#Dealbook 158.25144897 0.16623051 952
## ## 180.32449190 0.14043964 1284
## Metro#N.Y./Region# 30.35960093 0.15333132 198
## Culture#Arts# 85.20930589 0.12623601 675
## Business#BusinessDay#SmallBusiness 14.98537727 0.10703841 140
## Styles##Fashion 7.36986388 0.06193163 119
## #Opinion#RoomForDebate 6.46512194 0.10427616 62
## #Multimedia# 9.31950181 0.06609576 141
## Travel#Travel# 6.71907976 0.05742803 117
## TStyle## 31.23918133 0.04314804 724
## Foreign#World#AsiaPacific 13.72935437 0.06763229 203
## myOther 2.82545315 0.07435403 38
## Culture## 0.05194618 0.05194618 1
## #U.S.#Education 10.77780084 0.03316246 325
## Foreign#World# 6.29953592 0.03662521 172
## err.abs.new.sum err.abs.new.mean .n.new
## OpEd#Opinion# NA NA 164
## Styles#U.S.# NA NA 61
## Science#Health# NA NA 57
## Business#Crosswords/Games# NA NA 42
## #Opinion#ThePublicEditor NA NA 10
## Business#Technology# NA NA 114
## Business#BusinessDay#Dealbook NA NA 304
## ## NA NA 342
## Metro#N.Y./Region# NA NA 67
## Culture#Arts# NA NA 174
## Business#BusinessDay#SmallBusiness NA NA 41
## Styles##Fashion NA NA 15
## #Opinion#RoomForDebate NA NA 20
## #Multimedia# NA NA 52
## Travel#Travel# NA NA 35
## TStyle## NA NA 105
## Foreign#World#AsiaPacific NA NA 56
## myOther NA NA 5
## Culture## NA NA 70
## #U.S.#Education NA NA 89
## Foreign#World# NA NA 47
## .n.OOB .n.Fit .n.Tst .freqRatio.Fit
## 1728.000000 NA 1870.000000 NA
## .freqRatio.OOB .freqRatio.Tst err.abs.fit.sum err.abs.fit.mean
## 1.000000 1.000000 NA NA
## .n.fit err.abs.OOB.sum err.abs.OOB.mean err.abs.trn.sum
## NA 328.814611 4.170523 1037.439449
## err.abs.trn.mean .n.trn err.abs.new.sum err.abs.new.mean
## 3.432938 6532.000000 NA NA
## .n.new
## 1870.000000print(orderBy(as.formula(paste0("~ -", glb_sel_mdl_id, ".importance")),
subset(glb_featsimp_df, importance > 10)))## All.X.glmnet.importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 98.38173
## NDSSName.my.fctr#Opinion#ThePublicEditor 87.34197
## NDSSName.my.fctrScience#Health# 84.05064
## NDSSName.my.fctrStyles#U.S.# 80.66804
## NDSSName.my.fctrBusiness#Technology# 43.29623
## WordCount.log1p 34.01597
## WordCount.root2 32.29795
## .rnorm 31.33944
## NDSSName.my.fctrCulture## 31.33944
## NDSSName.my.fctrCulture#Arts# 31.33944
## NDSSName.my.fctrMetro#N.Y./Region# 31.33944
## WordCount.nexp 31.33944
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.33944
## NDSSName.my.fctrTravel#Travel# 31.31570
## NDSSName.my.fctrForeign#World# 30.00852
## NDSSName.my.fctr#Multimedia# 28.91940
## NDSSName.my.fctrmyOther 27.85141
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 27.78546
## NDSSName.my.fctrStyles##Fashion 22.02991
## NDSSName.my.fctrForeign#World#AsiaPacific 19.29994
## NDSSName.my.fctr#U.S.#Education 18.25900
## NDSSName.my.fctrTStyle## 17.36032
## importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 99.96082
## NDSSName.my.fctr#Opinion#ThePublicEditor 86.61487
## NDSSName.my.fctrScience#Health# 82.33065
## NDSSName.my.fctrStyles#U.S.# 77.12194
## NDSSName.my.fctrBusiness#Technology# 38.62774
## WordCount.log1p 37.41260
## WordCount.root2 32.88366
## .rnorm 31.93272
## NDSSName.my.fctrCulture## 31.93272
## NDSSName.my.fctrCulture#Arts# 31.93272
## NDSSName.my.fctrMetro#N.Y./Region# 31.93272
## WordCount.nexp 31.93272
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.12990
## NDSSName.my.fctrTravel#Travel# 29.62778
## NDSSName.my.fctrForeign#World# 25.48953
## NDSSName.my.fctr#Multimedia# 24.73162
## NDSSName.my.fctrmyOther 26.20874
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 23.59037
## NDSSName.my.fctrStyles##Fashion 21.10031
## NDSSName.my.fctrForeign#World#AsiaPacific 14.61070
## NDSSName.my.fctr#U.S.#Education 14.34221
## NDSSName.my.fctrTStyle## 16.67670
## Final.glmnet.importance
## NDSSName.my.fctrOpEd#Opinion# 100.00000
## NDSSName.my.fctrBusiness#Crosswords/Games# 99.96082
## NDSSName.my.fctr#Opinion#ThePublicEditor 86.61487
## NDSSName.my.fctrScience#Health# 82.33065
## NDSSName.my.fctrStyles#U.S.# 77.12194
## NDSSName.my.fctrBusiness#Technology# 38.62774
## WordCount.log1p 37.41260
## WordCount.root2 32.88366
## .rnorm 31.93272
## NDSSName.my.fctrCulture## 31.93272
## NDSSName.my.fctrCulture#Arts# 31.93272
## NDSSName.my.fctrMetro#N.Y./Region# 31.93272
## WordCount.nexp 31.93272
## NDSSName.my.fctrBusiness#BusinessDay#Dealbook 31.12990
## NDSSName.my.fctrTravel#Travel# 29.62778
## NDSSName.my.fctrForeign#World# 25.48953
## NDSSName.my.fctr#Multimedia# 24.73162
## NDSSName.my.fctrmyOther 26.20874
## NDSSName.my.fctrBusiness#BusinessDay#SmallBusiness 23.59037
## NDSSName.my.fctrStyles##Fashion 21.10031
## NDSSName.my.fctrForeign#World#AsiaPacific 14.61070
## NDSSName.my.fctr#U.S.#Education 14.34221
## NDSSName.my.fctrTStyle## 16.67670dsp_myCategory_conf_mtrx <- function(myCategory) {
print(sprintf("%s OOB::myCategory=%s confusion matrix & accuracy: ",
glb_sel_mdl_id, myCategory))
print(t(confusionMatrix(
glbObsOOB[glbObsOOB$myCategory == myCategory,
paste0(glb_rsp_var_out, glb_sel_mdl_id)],
glbObsOOB[glbObsOOB$myCategory == myCategory, glb_rsp_var])$table))
print(sum(glbObsOOB[glbObsOOB$myCategory == myCategory,
predct_accurate_var_name]) /
nrow(glbObsOOB[glbObsOOB$myCategory == myCategory, ]))
err_ids <- glbObsOOB[(glbObsOOB$myCategory == myCategory) &
(!glbObsOOB[, predct_accurate_var_name]), glb_id_var]
OOB_FNerr_df <- glbObsOOB[(glbObsOOB$UniqueID %in% err_ids) &
(glbObsOOB$Popular == 1),
c(
".clusterid",
"Popular", "Headline", "Snippet", "Abstract")]
print(sprintf("%s OOB::myCategory=%s FN errors: %d", glb_sel_mdl_id, myCategory,
nrow(OOB_FNerr_df)))
print(OOB_FNerr_df)
OOB_FPerr_df <- glbObsOOB[(glbObsOOB$UniqueID %in% err_ids) &
(glbObsOOB$Popular == 0),
c(
".clusterid",
"Popular", "Headline", "Snippet", "Abstract")]
print(sprintf("%s OOB::myCategory=%s FP errors: %d", glb_sel_mdl_id, myCategory,
nrow(OOB_FPerr_df)))
print(OOB_FPerr_df)
}
#dsp_myCategory_conf_mtrx(myCategory="OpEd#Opinion#")
#dsp_myCategory_conf_mtrx(myCategory="Business#Business Day#Dealbook")
#dsp_myCategory_conf_mtrx(myCategory="##")
# if (glb_is_classification) {
# print("FN_OOB_ids:")
# print(glbObsOOB[glbObsOOB$UniqueID %in% FN_OOB_ids,
# grep(glb_rsp_var, names(glbObsOOB), value=TRUE)])
# print(glbObsOOB[glbObsOOB$UniqueID %in% FN_OOB_ids,
# glbFeatsText])
# print(dsp_vctr <- colSums(glbObsOOB[glbObsOOB$UniqueID %in% FN_OOB_ids,
# setdiff(grep("[HSA].", names(glbObsOOB), value=TRUE),
# union(myfind_chr_cols_df(glbObsOOB),
# grep(".fctr", names(glbObsOOB), fixed=TRUE, value=TRUE)))]))
# }
print("glbObsNew prediction stats:")## [1] "glbObsNew prediction stats:"if (glb_is_regression)
print(myplot_histogram(glbObsNew, paste0(glb_rsp_var_out, glb_fin_mdl_id)))
if (glb_is_classification)
print(table(glbObsNew[, paste0(glb_rsp_var_out, glb_fin_mdl_id)]))##
## N Y
## 1144 726# Use this to see how prediction changes by changing one or more values
# players_df <- data.frame(id=c("Chavez", "Giambi", "Menechino", "Myers", "Pena"),
# OBP=c(0.338, 0.391, 0.369, 0.313, 0.361),
# SLG=c(0.540, 0.450, 0.374, 0.447, 0.500),
# cost=c(1400000, 1065000, 295000, 800000, 300000))
# players_df$RS.predict <- predict(glb_models_lst[[csm_mdl_id]], players_df)
# print(orderBy(~ -RS.predict, players_df))
# dsp_chisq.test(Headline.contains="[Vi]deo")
if ((length(diff <- setdiff(names(glbObsTrn), names(glbObsAll))) > 0) ||
(length(diff <- setdiff(names(glbObsFit), names(glbObsAll))) > 0) ||
(length(diff <- setdiff(names(glbObsOOB), names(glbObsAll))) > 0) ||
(length(diff <- setdiff(names(glbObsNew), names(glbObsAll))) > 0)) {
print(diff)
stop("glbObs* not in sync")
}
if (glb_save_envir)
save(glb_feats_df, glbObsAll,
#glbObsTrn, glbObsFit, glbObsOOB, glbObsNew,
glb_models_df, dsp_models_df, glb_models_lst, glb_model_type,
glb_sel_mdl, glb_sel_mdl_id,
glb_fin_mdl, glb_fin_mdl_id,
file=paste0(glb_out_pfx, "prdnew_dsk.RData"))
# tmp_replay_lst <- replay.petrisim(pn=glb_analytics_pn,
# replay.trans=(glb_analytics_avl_objs <- c(glb_analytics_avl_objs,
# "data.new.prediction")), flip_coord=TRUE)
# print(ggplot.petrinet(tmp_replay_lst[["pn"]]) + coord_flip())
glb_chunks_df <- myadd_chunk(glb_chunks_df, "display.session.info", major.inc=TRUE)## label step_major step_minor label_minor bgn end
## 16 predict.data.new 8 0 0 233.199 242.389
## 17 display.session.info 9 0 0 242.390 NA
## elapsed
## 16 9.19
## 17 NANull Hypothesis (\(\sf{H_{0}}\)): mpg is not impacted by am_fctr.
The variance by am_fctr appears to be independent. #{r q1, cache=FALSE} # print(t.test(subset(cars_df, am_fctr == "automatic")$mpg, # subset(cars_df, am_fctr == "manual")$mpg, # var.equal=FALSE)$conf) # We reject the null hypothesis i.e. we have evidence to conclude that am_fctr impacts mpg (95% confidence). Manual transmission is better for miles per gallon versus automatic transmission.
## label step_major step_minor label_minor bgn
## 10 fit.models 6 0 0 35.238
## 14 fit.data.training 7 0 0 181.344
## 11 fit.models 6 1 1 139.591
## 12 fit.models 6 2 2 161.330
## 15 fit.data.training 7 1 1 222.039
## 9 select.features 5 0 0 25.428
## 1 import.data 1 0 0 9.037
## 16 predict.data.new 8 0 0 233.199
## 13 fit.models 6 3 3 174.894
## 2 inspect.data 2 0 0 18.458
## 5 extract.features 3 0 0 22.532
## 8 partition.data.training 4 0 0 24.288
## 3 scrub.data 2 1 1 21.340
## 4 transform.data 2 2 2 22.441
## 6 manage.missing.data 3 1 1 24.178
## 7 cluster.data 3 2 2 24.259
## end elapsed duration
## 10 139.590 104.352 104.352
## 14 222.039 40.695 40.695
## 11 161.329 21.738 21.738
## 12 174.893 13.563 13.563
## 15 233.199 11.160 11.160
## 9 35.237 9.809 9.809
## 1 18.458 9.421 9.421
## 16 242.389 9.190 9.190
## 13 181.343 6.450 6.449
## 2 21.340 2.882 2.882
## 5 24.177 1.645 1.645
## 8 25.428 1.140 1.140
## 3 22.440 1.100 1.100
## 4 22.531 0.091 0.090
## 6 24.259 0.081 0.081
## 7 24.288 0.029 0.029
## [1] "Total Elapsed Time: 242.389 secs"